PB84-224468
Assessing the Releases and Costs
Associated with Truck Transport of
Hazardous Wastes
iCFf Inc., Washington, DC
Prepared for
Environmental Protection Agency, Washington, DC
1984
U.S. Depvtm* flf
rMINUUi IMHKS
-------
VBM-22A468
ASSESSING THE RELEASES AND COSTS ASSOCIATED
WITH TRUCK TRANSPORT OF HAZARDOUS WASTES
This report was prepared for
the Office of Solid Waste under
contract no. 68-01-6621
U.S. ENVIRONMENTAL PROTECTION AGENCY
Washington, D.C.
-------
This report was prepared by Dr. Mark Abkowitz and Dr. Amir Eiger, Faculty
Members, Department of Civil Engineering, Rensselaer Polytechnic Institute, Troy,
N.Y., and Mr. Suresh Srinivasan of Transportation Consultants, for the U.S.
Environmental Protection Agency and ICI* Incorporated under contract.
The report has been reviewed by the U.S. Environmental Protection Agency (EPA)
and approved for publication. Its publication does not signify that the contents
necessarily reflect the views and policies of the U.S. EPA, nor does mention of
commercial products constitute endorsement or recommendation for use by the U.S.
government.
-------
am. eti
noO n , DOCUMENTATION
PAGE I I
3. ! !:: :i Mssslen No.
PkRA 2244tR
4. ThΌ ad 3.MIfl.
Assessing the Releases and Costs Associated With
Truck Transport of Hazardous Wastes
S. upon Os.
1984
..
7. kiSatO
ICF incorporated
3. F,dpmdn Orpalallon Sept. No.
N/A
t. Petaks O,ponindoa Sin sad teas it PvWuet#TatWWo,t LISt No.
Office of Solid Waste (1*1-562)
US EPA
401 14 St. SW
Washington, D.C. 20460 68-01 -6621
mm
12. Sponsoiton Oig.nkmlon Name end MSsa 11. Type of Sepal I Petted Covered
Office of Solid Waste
Final
(sane) ______
14.
it Suppiemate.p Nets
IS. Mutton (USt SOS ati
This report estimates the releases from and costs of the truck transport of hazardous
waste. The report contains these estimates for bulk and container shipments. This
study is a component of a larger analysis of hazardous waste management, EPAs RCRA
Risk-Cost Analysis Model. 11 Transport releases are e esented as the sum of (1) the
expected fraction released enroute (ranging from 10 to lO per mile) and (2) the
expected fraction released at terminal points! (ranging from io to io4 per
shipment). The report estimates, usinq a cost formula, average costs of $4 to $5 per
mile, depend1 nn tvne of transoort. To make these estimates, .i& reviewed existing
studies and evaluated state and national data on accident rates, quantiti6; released
in accidents, distance of shipments, numbers of shipments, quantities shipped, and
component costs.
I?. Doeaneut AalyNe a. fl..al..iu.. .
C. COSATI flutdlOo.p
It ft atkbN
NTIS
S
I
S La - a
-------
ACKNOWLEDGEMENT
The authors would like to acknowledge the advice, guidance and
cooperation of Curtis Haymore, Arline Sheehan and Eric Males of the
Office of Solid Waste, U.S. Environmental Protection Agency. The
assistance provided by Joseph Kirk, Leslie Kostrich, Stephen Bailey
and Jean Tilly of ICF Incorporated is also sincerely appreciated.
Finally, substantial and useful comments during the review process
were made by Russell Cappelle of American Trucking Associations,
Inc., Joseph Nalevanko of the U.S. Department of Transporations
Materials Transportation Bureau and John Thompson of the Office of
Solid Waste, U.S. Environmental Protection Agency.
iii
-------
ACKNOWLEDGEMENT
TABLE OF CONTENTS
PAGE
I II
EXECUTIVE SUMMARY
Fraction Release Analysis Methodology .
Data Description
Estimating the Truck Accident Rate
Incident Modeling
Estimating the Expected Amount Released
Estimating the Cost of Transporting Waste.
Trip Profile Analysis.
Cost Methodology.
Model Application
Release Computation
Cost Analysis
Concluding Remarks .
3
4
6
7
9
10
10
12
14
14
15
15
CHAPTER 1 INTRODUCTION
CHAPTER 2 FRACTION RELEASE ANALYSIS METHODOLOGY
CHAPTER 3 DATA DESCRIPTION
3.1 Truck Accident and Volume Data .
3.1.1 Texas
3.1.2 California
3.1.3 New Jersey
3.2 Hazardous Waste Shipment Information
3.2.1 California
3.2.2 Texas
3.2.3 Massachusetts
3.2.4 New York
17
21
25
25
26
26
26
28
29
31
31
34
iv
-------
3.3 Hazardous Waste Incident Data.
CHAPTER 4 TRIP PROFILE ANALYSIS
4. 1 Data Refinement
4.2 Analyss Results
4.2.1 California
4.2.2 Texas
4.2.3 Massachusetts .
4.2.4 New vork
4.3 Implications of Pooling State
4,4 Summary
CHAPTER 5 INCIDENT MODELING . *
5. 1 Container Classification
5.2 Incident Occurrence Model
5.3 Estimating the Mean Shipment
5.4 Fraction Release Model.
5.5 Fraction Release Estimators
5.6 Fraction Release Estimates
5.7 Errors of the Estimates
5.8 Results and Implications
CHAPTER 6 ESTIMATING THE TRUCK
6.1 Analysis
34
40
40
42
42
45
48
51
54
56
58
59
61
Distance 65
67
76
77
79
81
ACCIDENT RATE 83
Data
84
V
-------
6.2 Results and Implications . 86
CHAPTER 7 ESTIMATING THE COST OF TRANSPORTING WASTE 93
7.1 Literature Review 93
7.2 Revised Procedure 104
7.2.1 Average Cost Approach - 6,000 Gallon Tanker. . . 106
7.2.2 Average Cost Approcah - 18 Ton Stake Truck . . 107
7.2.3 Deriving Cost Formulas 108
7.3 Comparison with Actual Charges 109
7.4 Summary lii
CI4APTER 8 MODEL APPLICATION AND CONCLUDING REMARKS 112
8.1 Scenario 1 112
8.1.1 Release Computation 112
8.1.2 Cost Analysis 114
8.2 Scenario 2 115
8.2.1 Release Computation 115
8.2.2 Cost Analysis 116
8.3 Concluding Remarks 116
REFERENCES 118
APPENDIX A LIST OF CONTAINER TYPES 121
APPENDIX B DESCRIPTION OF FAILURE MODES AND CAUSE
CODES 132
APPENDIX C INCIDENT FREQUENCY AND DAMAGE HISTOGRMS . 135
vi
-------
EXECUTIVE SUMMARY
In response to a growing concern over the management of
hazardous wastes and their impact on the population and envronment,
the Resource Conservation and Recovery Act (RCRA) was enacted in
1976. RCRA authorized the EPA to establish a hazardous waste
control program for the nation, which includes the identification and
classification of hazardous wastes. reements for owners and
operators of hazardous waste facilities, and guidelines for state
programs developed under the act.
In 1981. as part of the national hazards waste control program,
EPAs Office of Solid Waste began to develop its RCRA Risk/Cost
Analysis Model. The model is designed to assist in the development
of hazardous waste policies.
The RCRA Risk/Cost Analysis Model consists of an array of
possible ways to treat, transport and dispose of the hazardous wastes
9enerated in the United States. There are three main factors
considered in the models formulation of po .able says to n fl ge
hazardous waste:
(1) The type of waste (and its hazardous chemical
constituents).
(2) The types of technologies used to treat, transport and
dispose of the wastes.
(3) The environmental settings in which the wastes are
treated, transported and disposed.
-------
2
The model forms all possible combinations of a list of wastes,
technologies and environmental settings - - or W-E-T cells. The model
then calculates the risks and costs involved in each W-E-T cell. In
this fashion, the relative merits and drawbacks of various hazardous
waste management strategies can be identified.
This report focuses on one component of the RCRA Risk/Cost
t.nalysis Mode l: the costs incurred and expected fraction released
(Rtr) during transport of hazardous wastes. The objectives of our
project were governed by the following criteria:
In order to establish a tool for policy analysis, we wanted
to estimate a fraction release model that reflected, as much
as possible, actual data on hazardous waste shipments and
incidents. Compiling a comprehensive data sample
necessitated extensive data collection at both the state and
federal levels.
In order to ascertain whether previous studies were reliable
for policy analysis, we performed a critical review of
existing truck transport cost studies. We then developed
revised cost formulas to account for deficiencies identified in
the review process and compared the revised cost procedure
with quoted rates to validate its applicability.
Because 90 percent of all current hazardous waste transport is via
truck, the transport release model and cost review were restricted to
truck transport. 1
The authors are presently conducting studies of the release rates
and costs of hazardous waste shipments by rail and waterborne
transport.
-------
3
Fraction Release Analysis Methodology
Hazardous waste releases during transport can result from a
number of causes (failures modes) and can occur either at shipping
terminal points or enroute. We defined three incident types:
(1) Container failures due to vehicular accidents enrout .
(2) Container failures occurring enroute due to causes other
than vehicular accidents.
(3) Container failures at shipment terminal points.
We formulated a Transport Release Model to compute the expected
fraction released (Rtr) during transport. This is a function of: (1)
the expected fraction released enroute (2) the expected fraction
released at terminal points. Deriving these release fractions requires
an understanding of the expected fraction released given an incident
for each failure mode, the probability of an incident for each failure
mode and, for enroute incidents, the distance shipped. It is
necessary to estimate these parameters for each container type used
in transport. Thus, the total number of parameters to be estimated
depends on the number of container types and failure modes.
Furthermore, the use of the model for policy analysis requires
hazardous waste shipment distances as input.
Estimating incident probabilities also requires a determination of
the total involvement. For example, total involvement for incidents
which occur enroute is a function of the total distance shipped (i.e.,
the average shipment distance multiplied by the number of
shipments). For incidents which occur at terminal points, the total
-------
4
involvement is the total number of shipments. Thus, it is necessary
to estimate the average shipping distance arid the number of
shipments for each container type.
We computed these measures using: (1) shipping distances
derived from incident data, 2) data on the number of vehicular
accidents and 3) independently derived estiraates of vehicular accident
rates. Subsequently, it became possible to compute incident rates for
other failure modes. It was not necessary to perform this explicitly
for each container type. Rather, we expressed all incident rates in
terms of a common vehicle accident rate. We assumed that this
accident rate does not depend on the container type used for
shipment.
Data Description
We identified three types of data which were necessary to
conduct the release and cost analyses:
(1) Truck accident and volume data.
(2) Hazardous waste shipment iniformatron.
(3) Hazardous waste incident data.
Wherever possible, we obtained data from 1980, 1981 and 1982,
because they represent the most recent information available on
hazardous waste incidents and shipments.
We obtained truck accident and volume data from Texas,
California and New Jersey records. Each record included average
daily counts of vehicular traffic characterized by vehicle type and the
annual number of truck accidents. The California and Texas data
included observations for interstate highways, U.S. highways and
state routes. The New Jersey data, on the other hand, included
-------
5
many highway sections containing intersections with traffic signals.
We collected data on hazardous waste shipments from California,
Texas, Massachusetts and New York manifest records. In general,
each record contained the following intormation: origin location,
destination location, waste type transported, quantity shipped and
unit of shipment. A significant problem with this database was its
tack of accuracy in reporting The locations of generation and disposal
sites. In some cases, the County of or the destination state
was the only location description. Thus, it was necessary to make
some assumptions to correct for this problem. State data also did not
consistently include interstate shipments.
The primary data source for estimating the incident probability
and fraction release parameters was the Hazardous Material Incident
File (HAZMAT) maintained by the U.S. Department of Transportations
Materials Transportation Bureau (MTB). HAZMAT, a compilation of
naionw,de data on hazardous material spills, contains information on
the frequency and circumstances (container involverient, failure mode,
severity of result;ng spills, etc.) surroundirg te zardous material
incidents.
Although over 8,000 incidents of hazardous material spills
involving truck travel were repurted In 1981, a closer inspection of
these data indicated that an extremely small number (84) of these
spills involved hazardous wastes. Because the sample size of
hazardous waste incidents was not large enough for statistical
analysis, we considered all of these hazardous materials incidents in
developing the incident model. Also, because we postulated that the
incident rate and fraction release modek do not depend on the type
-------
6
of waste being shipped, but rather, on the container type used, and
because the HAZMAT file covers a wide range of container types, this
approach is justified.
Estimating the Truck Accid.nt Rat.
We assumed that the truck accident rate is a function of the
highway type and traffic conditions. Truck accident and volume data
were obtained from California, TeKas and New Jersey; these data
represented a wide range of traffic and truck volumes and four
different highway types. To test the statistical significance of any
differences in accident rates under different highway and traffic
conditions, we conducted an analysis of variance (ANOVA), which
indicated the significance of the traffic volume, truck percentages and
highway type.
The analysis of the accident rate data yielded the following
estimate for aggregate accident involvement rates (releasing accidents
per million truck mi!es):
Interstates 0.13
U.S. and State Highways 0.45
Urban 0.73
Composite 0.28
These results fall within the range of previously reported
estimates and demonstrate the difference in the accident rate for
various highway types. The truck accident rate is also dependent on
both the total traffic volume and the percentage of trucks in the
-------
7
traffic stream. These results suggest that in appying the estimates
provided, ceH means should be used in lieu of aggregate means if
sufficient information is available to identify the highway type and the
traffic volume.
Incident Modeling
The HAZMAT file of reported hazardous materials incidents
allows the coding of up to 334 container types and 27 failure modes.
From our anal ses of these dut , we oeu iried 8 container types with
reasonably uniform physical characteristics and incident involvement
rates:
(1) Cylinders
(2) Cans
(3) Glass
(4) PLastic
(5) Fiber Boxes
(6) Tanks
(7) Metal Drums/Pails
(8) Open Metal Containers
For each of these container classes, we determined the
respective parameters in the fraction release model. Table 1
summarizes the resulting estimates of the fraction released by
coniainer type.
The results of our analyses indicate that in terms of their
order of magnitude, the expected fractions released per mile shipped
range from io.. .8 to uc 6 , depending on the container class. The
sxpected fractions released at terminal points range from i a 6 to
depending on the container class.
-------
8
table 1 Estiutates of Fraction Released by Container Class
Container Expected Fraction Expected Fraction
Class Released Per Released at Terminal
Nile Shipped ** Poi nts
1 1.3 x 10_b +(.13 A) 1.4 * toe 4
2 2.6 x to 6 + (.12 A) 4.0 x L0
3 1.7 * 1o +C27 A) 2.6 x lO
4 6.1 x +(.14 A) 5.2 x 10
5 1.3 x ic ,a6 + .12 A) 6.1 x lO
6 4.2 x 10 (.19 A) 7.6 x 10°
7 2.4 1o +(.10 A) 2.9 x 1O
8* 7.5 x io 1.2 c
*astimate associated with the release fraction dining accident is not
reliable.
releasing vehicle accident rate.
-------
9
Our computed estimates indicate that:
(1) The release rates for tank trucks are much lower than
f3r other container types.
(2) The expected amount released at terminal points is one to
three orders of magnitude higher than the amount
released en route.
(3) The expected reida e during transport are
potentially as high as the release fractions at disposal
sites and treatment facilities which range from to
1O for routine spillage and 1O to 1O for accidental
spillage.
Estimating the Expected Amount Released
Using the model parameters given in the previous sections, we
employed the foltowin j procedure to estimate the expected fraction
released during transport:
(1) Identify shipment characteristics.
- number of shipments
- voiLme per shipment
- trip distance
- container type
(2) Identify highway characteristics.
- highway type
- traffic volumes
(3) Select appropriate values of fraction release parameters
for the container type being considered.
(4) Compute the fraction of accidents that involve releases
-------
10
(derived as the truck accident rate multiplied by 0.2).
(5) Determine fraction released enroute and at terminal
points.
(6) Multiply fraction released enroute by total trip miles and
fraction released at terminal points by the number of
shipments.
(7) Add these values to arrive at total expected fraction
released.
(8) Multiply this by the total volume to obtain the total
expected amount released.
This procedure is demonstrated in the discussion on model application.
Estimating the Cost of Transporting Waste
Trip Profil. Analysis
Using the waste shipment data from Texas, California,
Massachusetts and New York, we examined the following
(1) The mean sl ipping distance, segmented by waste type
(for each state).
(2) The quantity shipped, segmented by waste type (for
each state).
(3) The extent to which the above measures vary across
states.
The resulting information was used in cost applications where specific
trip lengths and the quantities shipped were not known.
In order to determine if the quantity and/or distance shipped is
related to the waste type (solid or liquid) or the particular state
under consideration, we conducted a multivariate analysis of variance.
-------
II
The results of the analysis indicated that the shipment characteristics
of liquid and solid wastes vary by state and consequently we could
not derive aggregate estimates. This resulted in our conducting
separate analyses for each state.
Our analysis results indicated that trip distance and quantity
shipped vary by waste category and also vary considerably among
statti. This is likely due to differences in the manifest system,
geographic location, size and industrial activity of each state.
We did, however, conclude that the quantity transported is
independent of trip distance. Our findings do not substantiate the
argument that shipments are filed closer to capacity on longer trips
than shorter ones. We also found that in three of the four states,
th. mean shipment size for liquids is dr9or than for solids
shipments, and that in three of the four states th. average trip
distance is longer for solids shipments than for liquids shipments.
Questions are sometimes raised regarding general waste shipment
characteristics for the United States. Although there is no basis for
assuming that our sample is typical of the tiie azardou. waste
transport industry, we computed weighted averages of the shipping
distances and quantities which reflect the number of annual manifests
in each of the states. These weighted averages should not be
misinterpreted to apply to specific hazardous waste transport scenarios
in the United States.
The mean trip length for all shipments is 84.2 miles, with a
mean trip length for liquids of 77.1 miles and for solids of 109.6
miles. For liquids, the mean quantity shipped is 3,171 gallons. For
solids, it is 2,791 gallons (11.6 tons). The trip distance frequency
-------
12
distribution for all four states for both liquids and solids follows an
exponential distribution. This is not surprising because disposal sites
ace hkely to be located near points of waste generation.
Cost Methodology
We reviewed the existing literature on the cost of transporting
hazardous waste and identified seven studies which treated the issue
of estimating the cost of transporting hazardous waste by truck. All
seven studios considered this issue within the larger framework of the
total cost and risk of hazardous waste treatment at a regional level.
The studies resuts varied from gross estimates of th. unit cost
of transport to more sophisticated derivations of costs ba.ed on fixed
and variable components. We noted several deficiencies in these
methods, part::ularly in the assumptions relating to shipment
characteristics (for example, all of the studies assumed that vehicles
travel at capacity, which is not substantiated by the results of the
trip profile analysis) and their failure to compare their results to the
actual rates charged by hawers.
Using the most comprehensive of the methodologies, we
developed a revised costing procedure which was designed to
overcome these deficiencies. Our modifications included considering
trip distances and shipment sizes based on the trip profile analysis
results, using 1983 component costs, and comparing the revised
methodology to actual price quotes from waste haulers.
We then used the revised costing procedure to estimate
transport costs for 6,000 gallon tankers and 18-ton stake (flatbed)
trucks. The average costs computed using the trip profile
characteristics are:
-------
13
Tankers Stake Trucks
Average Cost Per $4.14 $4.55
Loaded Mile (5)
Average Cost Per $0.31 $0.39
Loaded Ton-Mile (5)
The averag, costs per !oadsd mile and loaded ton-mile an, larger for
stake trucks than tankers. This is due to the smaller loads
aasocat.d with stake trucks.
In order to estimate the cost of transport when details on
specific siiipm.nts are available, we deri ed the following formulas for
tankers and stake trucks:
88.8
clmtank.r (S/loaded mile) 3.08 ____
3.08 88.8
cltm k (S/loaded ton-mile)
tan er
129.38
elmitak, (S/loaded mile) = 3.02
3.02 129.38
cltm k (S/loaded tori-m,le)
stae
where:
cim = cost per loaded mile
cltm = cost per loaded ton-mile
X shipment distance (miles)
V shipment size (tons)
-------
14
To determine th. accuracy of the revised costing procedure 1 wo
compared its estimates with the actual rates charged by haulers. The
comparison showed that the estimates we obtained using this cost
formula appear to be quite representative of quoted rates in the
hazardous waste transport industry. The average cost figures,
however, did not compare quite ai favorably. Consequently , we
r.comm.nd that the cost figures should be used rather
carefully . and should o iy be employed when information I. not
availabi , on distance and/or shipment size.
Modal Application
To illustrate th. established release and cost procedures, we
posed the following problem:
Supposs 200 55-gallon drums are being shipped a distance of 100
mites on interstate highways. The average daily traffic (ADT) and
truck percentages on the highways are unknown. What are the
.)pected releases and cost involved?
RIaM Computation
From previously reported results, we obtained the releasing
accident rate for interstates as 0.13 x io 6 releasing accidents per
truck mile. The expected amount released enroute was obtained using
the fraction released from Table 1 as:
E (release enroute) (2.4x10 6 b 0.lOxO.l3xU) 6 ) x 100 x 200 x 55
= 2.65 gallons
-------
15
E (release at terminals) = 2.9x10 4 x 200 x 55
3. 19 gallons
Total expected release = 5.84 gallons
Cost Analysis
The average load carried by stake trucks is 2,791 gallons
which is equivalent to 11.6 toni The quantity being shipped is
11.000 gallons. which is equivalent 45.83 tons. The cost per
loaded ton-mile is:
3.02 129.38
cltm (S/loaded ton-mile) 0.37
ate a 11.6 (100H11.6)
Number of ton-miles per shipment 11.6 x 100 1160
Cost per shipment 1160 0.37 $429.20
Average number of shipments = 3.94
Total Cost 3.94 429.20 = $1,691.05
Concluding Remarks
This project has addressed the potential releases and costs of
transporting hazardous wastes by truck. In the course of conducting
this study, we drew several conclusions that are useful for policy
analysis. Below, we briefly discuss our conclusions.
-------
16
A trip profile analysis conducted on data fror.i several states
indicated thdt, on average, wastes are shipped less than 100 miles
from their generation to their disposal sites. The average trip length
is lower for liquids than for solids. Generally speaking, the mean
quantity shipped i independent of shipping distance.
In assessing truck transport releases, it is important to
distinguish between two kinds of incidents that result in spills. For
one class of incidents, the probability of occurrence is a function f
the distance traveled; for the other, the occ rrence probability for a
particular shipment is fixed. We computed expected fraction release
estimates for both kinds of incidents.
The Costs of transporting hazardoub wastes by truck can be
reasonably approximated using the formulas derived in this study.
These cost formulas compare well with actual industry quotes.
The individual and collective results of the entire analysis are
applicable at many levels of aggregation. Using this studys models
and cott formulas, it is possible to obtain broad estimates of expected
releases and transport costs, as well as estimates of the releases and
costs involved in individual shipments.
Perhaps the most important result of this study is that the
releas. rates associated with transporting hazardous wastes by truck
appear to be as lar9e as the potential releases at treatment and
disposal sites. In fact, for some W-E-T combinations, transport may
be a potentially more dangerous activity. As a re:ult, policymakers
should give careful consideration to the relative risks involved in the
treatment, transport and disposal of hazardous wastes.
-------
17
CHAPTER 1
INTRODUCTION
In the United States, 160 million metric tons of hazardous
wastes are generated each year as part of the industrial process.
These wastes include organic chemicals, pesticides, acids, caustics,
flammables and explosives [ 11.
Accidents involving hazardous wastes have the potential to
produce catastrophIc effects on people and the environment.
Depending on the nature of the waste, the extent of Its release and
where it occurs, hazardous waste spills can impose serious public
safety problems through cont rninatIon of the surrounding air, water
or soil. Therefore, it is of utmost i np rnce to dispose of these
wastes with a minimal impact on the environment and to find safer
methods of transporting them from their generation zones to disposal
sites.
in response to a growing concern over the management of these
wastes and their impact n the population and environment, the
Resource Conservation and Recovery Act (RCRA) was enacted in
1976. RCRA authorized the EPA to establish a hazardous waste
control program for the nation, which includes the identification and
classification of hazardous wastes, requirements for owners and
operators of hazardous wasto facilities, and guidelines for state
programs developed under the act.
In 1981, as part of the national hazards waste control program,
EPAs Office of Solid Waste began to develop its RCRA Risk/Cost
-------
18
Analysis Model. The model is designed to assist in the development
of policies for hazardous waste facilities.
The RCRA Risk/Cost Analysis Model consists of an array of
possible ways to treat, transport and dispose of the hazardous wastes
generated in the United States (2]. There are three main factors
considered in the models formulation of possible ways to manage
hazardous waste:
(1) The type of waste (and its hazardous chemical
constituents).
(2) The types of technologies used to treat, transport and
dispose of the wastes.
(3) The environmental settings in which the wastes are
treated, transported and disposed.
Th ,odel forms all possible combinations of a list of wa it9s,
technologies and environmei tal settings. Thus, it may be regarded
as a three-dimensional matrix, each c .ll of which is a combination of
a waste, an environment and technology(ies) - - a W-E-T cell. Each
W-E-T cell may be viewed as a particular waste management practice.
The model then calculates the ri ks and costs involved in each
W-E-T cell. In this fashion, the relative merits and drawbacks of
various hazardous waste management strategies can be idsatified.
This report focuses on one component of the RCRA Risk/Cost
Analysis Model: the costs incurred and expected fraction released
(Rtr) during transport of hazardous wastes. The objectives of our
-------
1.9
project were governed by the following criteria:
In order to establish a tool for policy analysis, we wanted
to estimate a fraction release model that reflected, as much
as possible, actual data on hazardous waste shipments and
incidents. Cothpl ling a comprehensive data sample
necessitated extensive data cotR $on at both the state and
federal levels.
In order to ascertain whether previous studies were reliable
for policy analysis, we performed a critical review of
existing truck transport cost studies. We then developed
revised cost formulas to account for deficiencies identified in
the review process and compared the revised cost procedure
with quoted rates to validate its applicability.
8ecause 90 percent of all current hazardous waste transport is via
truck (3], the transport release model and cost review were
restricted to truck transport.
This report is organized as follows. Chapter 2 develops the
framework for the fraction release analysis and discusses the data
requirements. Chapter 3 summarizes the data collection effort and
describes the format of the database. Chapter 4 describes an
analysis of shipment characteristics performed on hazardous waste
manifest data from several states. Chapters 5 and 6 focus on the
estimation of the parameters for the fraction release model. Chapter
The authors are presently conducting studies of the release rates
and costs of hazardous waste shipments by rail and waterborne
transport.
-------
20
7 describes the procedure for estimating tne cost of transporting
wastes by truck. Chapter 8 provides examples demonstrating the use
of the fraction release and cost models, as well as some concluding
remarks. The appendices present the reports supporting
documentation.
-------
21
CHAPTER 2
FRACTION RELEASE ANALYSIS METHODOLOGY
Hazardous waste releases during transport can result from a
number of causes (failures modes) and can occur either at shipping
terminal points or enroute. Of those incidents which occur enroute,
a certain proportion results directly from truck accidents. We
defined three incident types as:
(1) Container failures due to vehicular accidents enroute.
(2) Container failures occurring enroute due to causes other
than vehicular accidents.
(3) Container failures at shipment terminal points.
In developing the transport release r.. -. ur postulates were made
for these three types of incidents:
(1) The probability of a truck accident in which a release
occurs is independent of the waste being shipped and
the container type used in shipment.
(2) The probabilit of occurrence of an incident at any point
along the route is a nonzero constant which, exclusive of
truck accidents, depends on the container type used.
(3) The probability of occurrence of an incident at a
shipping terminal point depends only on the container
type used.
(4) The expeηted amount released as the result of an
incident depends on the container type used and the
specific cause of the release (failure mode). It does not
-------
22
depend on the location of the incident.
We formulated the Transport Release Model as follows:
Rtr = R x A x d Expected fraction release enroute.
R x ! Expected fraction released at terminal points.
where: Rtr is the expected release fraction.
R is a vector of parameters corresponding to the
expected fraction released of hazardous wastes for each
defined failure mode.
A is the probability vector corresponding to incidents
enroute for each defined failure mode.
0 is the probability vector corresponding to incidents at
terminal points for each defined failure mode.
d is the distance shipped.
For each container type considered, it is necessary to estimate
the vectors R, 0 and A. Thus, the total number of parameters to be
estimated depends on the number of container types and defined
failure modes. Furthermore, the use of the model for policy analysis
requires hazardous waste shipment distances as input.
The primary data source for estimating the incident probability
and fraction release parameters in this analysis was the Hazardous
Material Incident File (HAZMAT) maintained by the U.S. Department
of Transportations Materials Transportation Bureau (MT B). A
compilation of nationwide data on hazardous material spills, HAZMAT
contains information relating to the frequency and circumstances
(container involvement, failure mode, etc.) surrounding hazardous
-------
23
material incidents.
Estimating incident probabilities also requires a determination of
the total involvement. For example, total involvement for incidents
which occur enroute is a function of the total distance shipped (i.e.,
the average shipment distance multiplied by the number of
shipments). For incidents which occur - terminal points, the total
involvement is the overall number of shipments. Thus, it is
necessary to estimate the average shipping distance and the number
of shipments for each container type.
The average shipping distance was computed from information
contained directly in the IIAZMAT file. We estimated the number of
shipments using 1) this estimate of the average shipping distance, 2)
HAZMAT data on the number of vehicular accidents and 3)
independently derived estimates of vehicular accident rates.
Subsequently, it became possible to compute incident rates for other
failure modes. It was not necessary to perTorm i. us explicitly for
each container type. Rather, we expressed all incident rates in
terms of a common truck accident rate. We assumed that this
accident rate does not depend on the container type used for
shipment.
After the Transport Release Models framework was developed,
we identified the following analyses and data requirements:
(1) Truck Accident and Volume Data
a. Compile truck accident rates for different highway types
and under different traffic volume conditions.
-------
24
b. Conduct statistical tests to determine the effect of highway
type, traffic volume and truck volume on the accident rate.
(2) Hazardous Waste Shipment Information
a. Compile average waste shipping distance and quantity
carried for several states and various waste categories
(solids, liquids, etc.).
b. Conduct statistical tests to determine the effects of states
and waste types on shipping distances and quantities.
(3) Hazardous Waste Incident Data
a. Identify container classes and failure modes to be
considered.
b. Estimate the mean shipping distances for each container
class.
c. Estimate the fraction released as a result of an incident for
each container class.
d. Estimate incident probabilities for each container class.
e. Derive expected release estimates for each container class
per mile shipped and at terminal points.
-------
25
CHAPTER 3
DATA DESCRIPTION
The previous discussion identified three streams of data which
were necessary to conduct the risk analysis:
(1) Truck accident and volume data.
(2) Hazardous waste shipment information.
(3) Hazardous waste incident data.
Wherever possible, we obtained data from 1980, 1981 and 1982,
because they represent the most recent information available on
hazardous waste Incidents and shipments. Below, we describe the
types and sources of the data gathered and the problems encountered
during data collection.
3.1 Truck Accident and Volume Data
To ensure that our database was comprehensive and useful, we
imposed the following rules for collecting truck accident and volume
data:
(1) Obtain a statistically-large sample of highway locations
for which accident histories, truck volumes and total
traffic volumes are available.
(2) Obtain location samplings for different highway types.
Different highway types (Interstate, U.S. routes, State,
etc.) are based on different design standards and, as a
result, may exhibit different accident frequencies.
(3) Obtain location samplings from several states. While the
-------
26
design standards are essentially the same across states,
there may be other variables which affect truck accident
rates (e.g., climate).
Following these rules, we obtained accident and volume data over
5-mile sections from three states: Texas, California and New Jersey.
3.1.1 Texas
The Texas State Department of Highways and Public
Transportation maintains 320 manual traffic volume count stations
across the state. These stations provide average daily counts of
vehicular traffic characterized by vehicle type. The Department also
maintains a comprehensive accident records system from which one can
obtain accident data. We obtained accident and volume data from 47
randomly selected stations (9 State, IS U. S. Routes and 20
Interstates) for the year 1980; the format of these data is described
in Table 3.1.
3.1.2 California
The California Department of Highways and Public
Transportation maintains count station data in the same basic format
as the Texas data. We randomly selected 95 count stations (46 State,
15 U.S. Routes and 34 Interstates) for the year 1981, and obtained
bi-directional volume and accident data for 5-mile sections for each
station.
3.1.3 New Jersey
The New Jersey Department of Transportation maintains
classified traffic counts as well as descriptions of vehicular accidents.
-------
27
Tabli 3.1 Truck Accident and Volume Data Format
DATA ITEM TYPE
Station cod. Text
Station Location Text
Highway Text
Control section Text
On.-dlrsctionai .ngth Real
Numb.r of truck accident/year Integer
Truck Average Daily freffic Integer
(ADT) ( 2 -ax le and greater)
Total ADT Integer
-------
28
We obtained data from 52 out of 171 randomly selected count stations
for 1980. The traffic volume counts from these stations were 8-hour
averages for 1980. The data were not segmented by highway type
because most of the sampled roadway sections contained signalized
intersections. Instead, the number of intersections in the 5-mile
segment of interest was recorded because we felt that intersections
would influence the accident rate more strongly than highway type.
3.2 Hazardous Waste Shipment Information
Some states have implememented a manifest system for recording
hazardous waste shipments. The data from such manifest systems can
be used to study the quantity and distance shipped by waste
category. We obtained waste shipment data from California, Texas,
Massachusetts and New York. We selected these states because they
organized and maintain accessible manifest records, and the states
vary in geographical location, size and in their level of industrial
activity.
it should be noted, however, that none of these states records
information on shipping modes as part of the manifest file. In order
to determine which shipments were made by truck, we assumed that a
maximum shipment weight of 66,500 lbs was transportable by truck,
based on information in the Oglesby and Hicks study (4].
2 Data availability imposed limitations on this analysis such that
different states were used for accident and hazardous waste shipment
analyses, respectively.
-------
29
3.2.1 CalifornIa
We obtained the entire manifest file from the California
Department of Health Services (CDHS) in two formats, A and B.
The major difference between the two formats is that Format A used
an alphanumeric code for the county where the shipment ortglnated
while Format B used a numeric code for the same information. The
CDHS used Format A through March 1981 and Format B from April
1981 through June 1981.
The CDHS also made a reporting change in its waste code.
Prior to Febrary 1981, it defined 16 waste types; the code was
subsequently expanded to include 76 waste types. The reporting
format allows for the possibility of three different waste types being
shipped concurrently, but the details of the shipment refer only to
the first waste code noted on the record. The general format of
these shipment data is shown in Table 3.2.
The manifest identifies each shipments point 3f origin by
county, implying that any analysis of shipments would be conducted
by assuming that travel originated at the countys centroid. The
disposal sites are identified by name and location. The data also
include out-of-state shipments, which comprise approximately 2 of all
shipments. Although the data do not specify the destination state, a
CDHS official estimated that 80 percent of interstate shipments are
destined for Nevada.
-------
30
Table 3.2 California Waste Shipment Data Format
DATA ITEM TYPE LENGTH
County code Integer 2
Waste type codes integer 6
Hazardous properties code Integer I
Number of containers Integer 3
Container type code tnteger I
Physical state of materials code Integer 1
Disposal site code Integer 3
Quantity shipped integer 5
Units integer 1
Handling method code Integer 1
Disposal date Integer 5
-------
3].
3.2.2 Texas
We obtained the entire manifest file for the years 1976, 1979,
1980 and 1981 from the Texas Department of Water Resources. 3
Initially this database included all hazardous material shipments, but
was subsequently modified to include only those shipments of materials
which are categorized as wastes. Because the database contained
very limited information regarding out-of-state disposal sites, we used
only the data on disposal trips within Texas. In cases where both
the shipper and receiver filed reports for the same shipment, we
eliminated these duplicate records.
Each record contains the registration numbers of the shipper
and receiver. We then used the master file of shippers and disposal
sites to obtain the exact origin and destination location of each waste
shipment. This allowed for us to make more accurate estimates of the
distance traveled than in California. Table 3.3 provides a description
of the Texas data format.
3.2.3 Massachusetts
We obtained a random sample of waste shipments for 1981 from
Urban Systems Research and Engineering, Inc., the firm responsible
for collecting hazardous waste data for the Massachusetts Bureau of
Solid Waste Disposal. The sample consisted of 642 records, which
includes both intrastate and interstate shipments. These data are
described in Table 3.4.
Although the origin of a shipment was identified by community,
its destination site was coded only by state. We obtained a separate
3 We requested the 1976 data in case it was necessary to perform a
trend analysis.
-------
32
Table 3.3 Texas Waste Shipment Data Format
DATA ITEM TYPE LENGTH
Report date Integer 4
Receiver district code Integer 2
Receiver registration number Text 5
Shipper district code Integer 2
Shipper registration number Text 5
Ticket number Integer 6
Ticket type Text 2
Waste code Integer 6
Quantity shipped Integer 8
Units code Integer 1
Date shipped integer 4
Comments Text 30
Record number Text 30
-------
33
Table 3.4 Massachusetts Waste Shipment Data Format
DATA ITEM TYPE LENGTh
Text 5
Name Text 25
wn of origin coc e Integer 3
Region of origin cod. Integer 1
Month in 1981 Integer 2
Waste type Text 2
Destination code Integer 2
Method of disposal code Integer 1
Generator code Integer 4
Employment at generation site Integer 7
Volume shipped (gals) Integer 6
-------
34
list of disposal sites which specifies each facilitys exact location and
the types of wastes it treats [ 5]. We assumed that all shipments
would go to the nearest facility in the destination state that could
accomodate the type of waste being transported.
3.2.4 New York
The New York data consist of a random sample of 209 records
foi 1982 that were randomly selected from a file of hazardous waste
shipments maintained by the New York State Department of
Environmental Conservation. These data include the town of origin
and destination for both intrastate and interstate shipments. The
waste shipment data format for New York appears in Table 3.5.
3.3 Hazardous Waste Incident Data
The U.S. Department of Transportations Materials
Transportation Bureau (MTB) collects information on hazardous
materials spills from all states. We obtained their entire data file
(called HAZMAT) from the National Data Corporation for the years
1976, 1980 and 1981. It should be noted that the MTB redefined the
term incident in January 1981 to exclude all battery spills and spills
of paints contained in 5-gallon cans or less, unless death, injuy or
excessive damage occurred.
The HAZMAT data allow for two container types to be coded for
each shipment. Container type 1 is usually the inner container and
container type 2 is the outer container (unless two different container
types are used in the same shipment). Failure modes are used to
-------
35
Table 3.5 New York Waste Shipment Data Format
DATA ITEM TYPE
Origin (town, stat.) Text
Destination (town, state) Text
Waste type Integer
Quantity Integer
Units Integer
Number of containers Integer
Container type 1 nteger
-------
36
describe the reasons for container failure (see Appendix B).
HAZMAT allows two such modes for each container type (e.g.,
handling failure and loose valves). An example of each record and
the Information it can contain is shown in Table 3.6.
Although over 8,000 incidents of hazardous material spills
involving road travel were reported In 1981, a closer inspection of
the data indicated that only 84 of these spills involved hazardous
wastes. Because the sample size of hazardous waste incidents was
not large enough for statistical analysis, we considered all hazardous
materials incidents in developing the incident model. In view of the
postulates made in Chapter 2 (i.e., that the incident rate and
fraction release models do not depend on the type of waste being
shipped, but rather, on the container type used), and the fact that
the HAZMAT file covers a wide range of container types, this
approach is justified.
Based on the classification of hazardous wastes used by the Materials
Transportation Bureau.
-------
37
Table 3.6 Hazardous Waste Incident Data Format
DATA ITEM TYPE LENGTh
Report number Text 8
Multiple code Text 1
Mode code Text 1
Date of incident Text 6
Time of incident Text 4
Incident city Text 13
Incident state Text 2
Carriers Text 9
Shippers Text 9
Origin city Text 13
Origin state Text 2
Destination city Text 13
Destination state Text 2
Injuries Integer 4
Deaths Integer 3
Damages Integer 8
Damage code Text 1
Quantity released Integer 7
Units Text 3
-------
38
Commodity code Text 5
Commodity class Text 2
Container 1 code Text 8
Failure code 1 cont 1 Integer 2
Failure coda 2 cont 1 Integer 2
Capacity container 1 Integer 6
Capacity units cent 1 Text 3
Number in shipment cent 1 Integer 5
Number failed cent 1 Integer 5
Gauge of cant 1 Text 6
Manufacturers of cont 1 Text 9
Label or placard Text 7
Completeness code Text 1
Sigr ficance of report Text 1
General cause of incident Text 1
R ssult of release Text 1
Recommendation on report Text 1
Apparent violation Text 1
Miscellaneous information Text 2
Container 2 code Text 8
Failure code 1 cont 2 Integer 2
Failure code 2 cont 2 Integer 2
Capacity container 2 Integer 6
Capacity units cont 2 Text 3
Number in shipment cont 2 Integer 5
-------
Number failed cont 2 Integer 5
Gauge of cont 2 Text 6
Manufacturers of cont 2 Text 9
Rail-tank-car ID no. Text 10
Registration exemption no. Text 6
Inspection date Text 6
Carriers name Text 30
Shippers name Text 30
COmmOdity flU1IS Text 19
39
-------
40
CHAPTER 4
TRIP PROFILE ANALYSIS
Using the waste shipment data from Texas, California,
Massachusetts and New York, we compiled the following:
(1) The mean shipping distance, segmented by waste type
(for each state).
(2) The quantity shipped, segmented by waste type (for
each state).
(3) The extent to which the above measures vary across
states.
The resulting information was used in cost applications where specific
trip lengths and the quantities shipped were not known. It also
serves as useful information for policy studies which rely on
characteristics of hazardous waste shipments. Below, we describe the
process used to refine the database and the analysis procedure for
each of the four state databases.
4.1 Data Refinement
After we eliminated records of non-hazardous waste shipments,
redundancies and other reporting problems, we were left with 56,414
records for Texas (1981), 40,245 records for California (1981), and
random samples of 642 records for Massachusetts (1981) and 209
records from New York (1982). For every state, we performed a
sampling procedure; the sample size was such that the 95 percent
confidence limits were within 30 percent of the mean (with the
-------
41
exception of the Massachusetts solids data).
For Texas, 137 records were randomly selected from the
database and for each of these records we identified the registration
of the shipper and receiver. We then obtained locations of the
generation and disposal sites using the master registration file of the
Texas Department of Water Resources. Finally, we used a road map
to estimate trip distances.
For California, we randomly selected 242 records using a similar
sampling scheme. Each record contained information on the origin
county (generation site) and the disposaf site locations Using road
maps, we identified county centroids and estimated the trip distance
from the origin centroid to the disposal site.
For New York, 193 randomly selected records out of the 209
were used in the analysis. A random sample of 233 Massachusetts
records were selected based on the sampling scheme.
If the generation and disposal sites were located in the same
town, we assumed a shipping distance of 10 miles for Texas,
California and New York. For Massachusetts, we assumed a 5 mile
shipping distance, as towns were assumed to be geographically smaller
there.
Shipments of solids in Massachusetts comprised too small a share of
overall shipments in the random sample to meet this criteria.
-------
42
4.2 Analysis Results
In order to determine if the quantity and/or distance shipped is
related to the waste type (solid or liquid) or the particular state
under consideration, we conducted a multivariate analysis of variance.
The results of the analysis indicated that the shipment characteristics
of liquid and solid wastes vary by state and consequently we could
not derive aggregate estimates. This resulted in our conducting
separate analyses for each state, as described below.
4.2.1 California
The California data on the quantity of waste shipped are coded
in five different units:
(1) Gallons.
(2) 42 gallon barrels.
(3) 55 gallon drums.
(4) Tons.
(5) Cubic yards.
We assumed that the first three codes cqnstitute a liquid measure,
while the last two are for sohds.
Figure 4.1 shows the shipping distance distribution for the
overall sample. Table 4.1 displays the means and standard deviations
of the shipping quantities and distances by waste type. The mean
shipping distance is roughly 78 miles, with liquids being transported
greater distances than solids. The latter was confirmed by a
hypothesis test which was significant at the 95 percent confidence
For states where both unit codes and waste type codes were
available, consistency checks were administered.
-------
43
Figure 4.1
2
a
2
Frequency Histogram for Overall Sample -
California
U.
10
0
100 00
0Z37 NC& (JIJl. 3I
-------
44
Table 4.1 Distance and Quantity Shipped - California
Dh tance Shipped
Waste Type Sample Mean St. Dev.
Liquid 77 99.08 68.02
Solid 165 68.39 94.00
Grand Mean 242 78.16 87.62
Quantity Shipped
Waste Type Sample Mea. St.Dev,
Liquid 77 3156. 1719.
Solid 165 2199. 1639.
Grand Mean 242 2504. 1720.
-------
45
level.
While it has been argued that haulers operate closer to capacity
h ng-distance runs, we found that the distance and quantities
shipped are uncorrelated (p 0.15).
On the basis of the above observations, one can conclude that,
in California, shipments involving liquids travel significantly greater
distances than those involving solids. Furthermore, the quantity of
waste shipped is, on the average, the same for varying trip lengths.
This is true both for liquids and solids.
4.2.2 Texas
The Texas data on the quantity of waste shipped are coded in
the following units:
(1) Tons.
(2) Gallons.
(3) Cubic yards.
(4) 55 gallon drums.
As before, we assumed that the tons and cubic yards codes
constitute a solids measure and that gallons and 55 gallon drums are
for liquids. Table 4.2 displays the means and standard deviations of
the shipping distances and quantities by waste type. Figure 4.2
shows the shipping distance distribution for the overall sample, which
again follows an exponential form.
The mean shipping distance in Texas is approximately 57 miles.
roughly 27 percent less than in California. It is interesting to note
that in Texas, solids shipments travel longer distances than liquids, a
-------
46
Table 4.2 DIstance and Quantity Shipped - Texas
Distance Shipped
Waste Type Sample Mean St.Dev.
Liquid 89 49.58 78.65
Solid 48 70.37 82.67
Grand Mean 137 56.87 80.39
Quantity Shipped
Waste Type Sample Mean St.Dev.
Liquid 89 3650. 1812.
SolId 48 3390. 2041.
Grand Mean 137 3481. 1961.
-------
47
Figure 4.2 Frequency Histogram for Overafl Sampe -
Texas
* so
0
Is
I -
a
U
S
w
D. Dr1r-
ill 10D 10 5 . 117 1
10 *10 110 3*0 ISO
D13T 1SCC UIZLC3*
-------
48
reversal from the California findings. This was confirmed by a
hypothesis test at the 95 percent significance level. We also found in
Texas that distance and quantities shipped are uncorrelated (p =
0.23).
4.2.3 Massachusetts
The Massachusetts data on waste types are coded in the
following units:
( ) Liquids (in gallons).
(2) Solids (in gallons).
Within these broad categories, waste types are coded by the nature
of the waste (solvents, waste oils, etc.) Figure 4.3 shows the
frequency histogram of shipping distances for the overall sample and
Table 4.3 displays a summary of the distance and quantity shipped
by waste type. Note that the mean trip length for all shipments is
similar to those for Texas (57 miles) and California (78 miles), states
which are much larger in size. The reason for this, of course, is
that the Massachusetts data reflects the fact that approximately 25% of
the waste is shipped out-of-state. As mentioned previously, the
California and Texas manifest data is primarily for within-state
shipments.
On the basis of the computed correlation between distance and
quantity siipped 1 the quantity of liquids shipped appears to be
independent of distance (p .27), whereas the quantity of solids
shipped is related to shipping distance (p .69). This find ng is
different from the California and Texas solids results. However, the
-------
49
Fl gure 4.3 Frequency Histogram for Overall Sample -
Massachusetts
Is.
a.
I,
*
S
to
_ nnn
1 50 ft. l173
10 110 110 U0 110
GiS RNC5 tJI1LL3I
-------
50
Table 4.3 Distance and Quantity Shipped - Massachusetts
Distance Shipped
Waste Type Sample Mean St.Dev.
Liquid 203 65.54 112.0
Solid 30 102.7 181.2
Grand Mean 233 70.32 123.2
Quantity Shipped
Waste Type Sample Mean St. Day.
Liquid 203 1438. 1769.
Solid 30 1009. 1495.
Grand Mean 233 1383. 1739.
-------
51
solids database is relatively small in Massachusetts, and the results
must be interpreted accordingly.
4.2.4 New York
The New York data on hazardous waste shipments are coded in
the following units:
(1) CubIc yards.
(2) Tons.
(3) Gallons.
Again, we assumed that the first two measures are for solids and the
third is for liquids. These data also indude several records showing
that wastes were either shipped out-of-state, or orginated in other
states but were disposed of in New York. We included these data in
the analysis.
Figure 4.4 shows the frequency histogram of shipping distances
for the overall sample and Table 4.4 displays a summary of the
distances and quantities shipped by waste type. As can be seen
from the table, New York has the longest mean shipping distance of
the four states (128 miles), and solids have a much longer mean
shipping distance than liquids. The latter observation was
substantiated by the hypothesis test which was significant at the 95
percent level. This is not surprising, as the New York data includes
interstate shipments, and solids are often transported long distances
to landfills while liquids often travel locally to recyclers or
incinerators. As in the case of the other states that were analyzed,
the quantities and distances shipped are uncorrelated (p 0.17).
-------
52
Ficiure 4.4 Frequency IHstograin for Overall Sample -
New York
a.
a.
0
U
a
1
SI.
o IJr _ ,r _ ,p __ ,r _ .1r _ I1r __ 1r! ,
iDa zoo soo oo 7S
a. 350 i SO
O13?R1I 1$JLLSI
-------
53
Table 4.4 Dlstanc. and QL antIty Shipped - New York
Dlstanc Shlpp d
Waste TYp S pip . Mean St.D.v.
Liquid lap 94.51 119.6
Solid 196.8 182.4
Grand Mean 19 127.9 150.5
Quantity Shipped
Waste Type Sample Mean St. Dcv.
Liquid 130 2972. 2143.
Solid 63 2968. 1894.
Grand Mean 193 2971. 2060.
-------
54
4.3 Implications of Pooling State Data
The results of the four state analyses varied both in terms of
distances traveled and quantities shipped. This is likely due to
substantial differences In the manifest system, location, size and
industrial characteristics of these four states. It was also shown that
there is no valid statistical argument for pooling the data from each
state into an aggregate sample.
However, questions are often raised regarding general waste
shipment characteristics for the United States. Although there is no
basis for concluding that our sample is typical of the hazardous waste
transport industry, we computed weighted averages of the shipping
distances and quantities which reflect the nuntber of annual manifests
in each of the states. These weighted averages should not be
misinterpreted to apply tospecific hazardous waste transport scenarios
in the United States.
The results appear in Table 4.5. The mean trip length for all
shipments is 84.2 miles, with a mean trip length for liquids of 77.1
miles and for solids of 109.6 miles. For liquids, the mian quantity
shipped is 3,171 gallons. For the solids categories, it is 2,791
gallons (11.6 tons). For both solids and liquids, the quantity
shipped increases slightly with trip length, but not enough to
support a statistically-significant conclusion, even at the 90 percent
confidence level.
-------
55
Table 4.5 DIstance and Quantity Shipped - Weighted Sample
Distance Shipped
Waste Type Sample Mean
Liquid 499 77.1
Solid 306 109.6
Grand Mean 805 84.2
Quantity Shipped
Waste Type Sample Mean
Liquid 499 3171.
Solid 306 2791.
Grand Mean 805 2931.
-------
56
The trip distance frequency distribution for all states, for both
liquids and solids, follows an exponential distribution. This is not
surprising because disposal sites are likely to be locatec near points
of waste generation. Thus, one can represent the distance
distribution as:
f(x) A e A
where:
x = shipping distance.
1/A mean of the distribution.
4.4 Summary
Using manifest data from California, Texas, New York and
Massachusetts, we examined two waste characteristics: trip distance
and quantity transported per shipment. Our analyses of these data
indicated that trip distance and quantity shipped vary by waste
category and also vary considerably among states. This is likely due
to differences in manifest systems, geographic location, size and
industrial activity of each state.
We concluded that the quantity transported is independent of
trip distance. Our findings do not substantiate the argument that
shipments are filled closer to capacity on longer trips than shorter
ones. We also found that in three of the four states, the mean
shipment size for liquids is larger than for solids shipments, and that
in three of the four states, the average trip distance is longer for
-------
57
solids shipments than for liquids shipments. Finally, we found that
shipping distance, in general, can be approximated by an exponential
distribution.
-------
58
CHAPTER 5
INCIDENT MODELING
As defined in Chapter 2, the three types of incidents which
result in the release of hazardous materials are:
(1) Container failures due to vehicular accidents enroute.
(2) Container failures occurring enroute due to causes other
than vehicular accidents.
(3) Container failures at shipping terminal points.
These incidents can result from a number of failure modes. We
assumed that the probability of an incident occurring depends on the
particular container used in shipment. In this chapter, we describe
the development of two models, the incident occurrence model and the
fraction release model (the fraction release model contains two
submodels, the fraction of containers failed and the fraction spilled).
From these models, we derived estimates for the expected fraction
released enroute and at terminal points for each of the identified
container classes.
ln the course of our analysis, we reviewed several studies
which also examined the risk of transporting hazardous materials (for
an overview discussion of this topic see TRB [ 6] and NCHRP(7]).
In general, the methodologies for determining estimates of risk can be
grouped in three broad categories: statistical estimation, fault-tree
analysis and subjective estimation. Each of these techniques has
advantages and disadvantages which must be evaluated in any given
case. For example, the primary limitation of statistical estimation
-------
59
techniques is the fact that one must assume the process generating
the accident/incident frequencies to be stationary. Otherwise, the
estimates obtained from past data could not be used to predict future
occurrences. Unlike statistical estimation methods, fault-tree analysis
attempts to model the incident occurrence process in great detail.
While this has great scientific appeal, there are difficulties associated
with the acquisition of data for predicting basic event probabilities
and the uncertainty that all significant event sequences have been
considered. Nevertheless, fault-tree analysis as applied to the
estimation of the risk of transporting h :ardous materials has been
used in several studies among which are Rhoads (8], Bercha (9] and
Geffen [ 10]. Other studies relevant to the evaluation of risk in
hazardous material transport include those of Gaylor (113, Jones (12]
and NTSB [ 13]. The reader is referred to a comprehensive
bibliography on this subject provided by Russell, et al. [ 143. Of
the various techniques discussed in the literature we considered
statistical estimation to be the most appropriate for the present study
in terms of the overall project objectives. We used the results of
other researchers to check the credibility of our estimates.
5.1 Container Classification
The HAZMAT file allows the coding of up to 334 container
types, 27 failure modes and 4 cause codes (see Appendices A and
B). We chose cause code 3, vehicular accident, to compute the
frequency of such accidents. This was done in order to avoid the
-------
60
possible ambiguity resulting when several failure modes appear in a
given record. The other three cause codes were considered too
general for this analysis and we discarded them in favor of the more
detailed failure modes.
We reduced the 334 container types to 42 by eliminating those
which had a low frequency of incidents (less than 10) during the
analysis year (1981). We then grouped the remaining container types
into 9 classes on the following basis:
(1) Similarity of physical characteristics (e.g., strength).
(2) Incident involvement.
A further analysis of the HAZMAT data revealed some records
with improperly coded container capacities and others with a mismatch
in the units for the quantities shipped and spilled. After we
eliminated these records, there were no observations in one of the 9
classes, so we eliminated that class. In addition, there were no
recorded observations for failure m ,des 23, 24, 25 and 26 in any
class. By eliminating these 4 modes, we were left with 23 failure
modes for analysis.
For each of the 8 remaining container classes we derived and
plotted incident frequency and damage histograms. These histograms
demonstrated that, in addition to their physical differences, the
container classes differed in terms of both failure frequency and
associated damage for the 23 failure modes. As a result of this step
in the analysis, we identified an additional container class. The final
list of container classes and tP e container types that comprise them is
-------
61
shown i i i Table 5.1. Th frequency and damage histograms for the
first 8 container classes (excluding the other class) are shown in
Appendix C.
5.2 Incident Occuirιhb _____
In order to i tIthit I,iobabiiities of failure from a database
containing frequeiilil i of ililurι, one requires a measure of the total
involvement. it i i4 b shoWn that if one assumes that the
probability of an iheident 1 constant along all points on a 9iven
route, then the probability of occurrrnc of an incident somewhere
along the route is directly roport onal to the length of the route.
Thus, for the first two incident types (incidents enroute), the total
transport distance is the total involvement. For incidents at shipment
terminal points, the number of shipments is the total involvement
since distance is not a. factor in this case. Given the above
conditions for each container class and failure mode, the limiting
probability distribution for the number of incidents is a Poisson
distribution. We demonstrate this result below for the number of
container failures occurring enroute by failure mode f for a
particular cbntainer class:
Let:
S be the number of shipments
F(d) be the cumulative probability distribution for the shipment
distances for the container class being considered
-------
62
Table 5.1 ContaIner Classification
Container Class Container Types
1. Cylinders 278,279
2. Cans 264,266,268
3. Glass 257,274,292,295
4. Plastic 258,276,296,320
5. Fiber Boxes 69,7,260,281
8. Tanks 2,,-iu,. .07. 308,309,310,312,313
315,322,327,328
7. Metal Drums/Pails 91,92,95,160,161,162,282
8. Open Metal Containers 318,319
9. Other 271,273,321,326
-------
63
be the mean distance shipped
N 1 be the number of incidents occurring en route by failure
mode i
p 3 be he r 0 9 of incident involvement by failure mode
j while enrou e
be the probability of incident involvement by failure mode
r pnit distance traveled
Then for a shiPmert length d, the probability of an incident by
failure node 3 p = X d. Furthermore, the total number of
shipments of distance d is S dF(d).
The random variable follows a binomial distribution with parameters
S dF(d) and
P [ NJ = n Id = binomial(S dF(dL 3
The binomial distribution can be approximated by a Poisson probability
mass function with parameter (p $ dF(d)):
P [ N 1 = n 3 jdJ Poisson [ S dF(d)x 1 d)
Using the result that the sum of independent Poisson random
variables is also a Poisson random variable with a parameter equal to
the sum of the individual parameters, we obtained:
P(N 1 = n 1 ] Poisson(E $ dF(d) Jd]
Poisson [ S
The same derivation can be used for each of the other incident
-------
64
types.
Thus, corresponding to each container class there are a set of
probability mass functions for the various incident types and failure
modes given by:
exp(-} dS) ()PdS)l
Container Failure P(nl S,A ,iid} - (1)
during Vehicular n 1 1
Accidents n.
exP(_ i dS)( JPdS) j
Container Failures P(n.lS,).. ,ILd) = J=2,23
Enroute
J (2)
Failures at P(m.IS,O.) = i = 1,23 (3)
Terminal Points m !
where S is the number of shipments, d is the mean shipping
distance, and the Xs and 0s are the corresponding incident rates.
We derivec 1 the estimators of ) and as:
1111 -
X j2, ...,23 (4)
ni
m 4l
0. X d j = 1,2, ..., 23 (5)
J ni
where X is an estimate of the truck accident rate in which releases
occur 2.8 x 10 from Section 5.6), d is an estimate of d the
mean shipping distance for the container class, to be determined from
the HAZMAT file (see Section 5.3), and and are the incident
frequencies for the container class (obtained from the HAZMAT file).
-------
65
Note that in equations 4 and 5, and 8 do not exist if
= 0. When tlils occurs, the effect of the Poisson approximation in
the derivation of the probability mass functions (equations 1,2, and
3) must be considered explicitly. The re5ulting estimators become:
(n 1 1)v exp(uN
= E 1 (uN) (6)
d
(m 1 41) u exp(uN) E 1 (uN) (7)
where N is the total number of observed incidents for the particular
container class, u Is computed by:
1 1
(8)
and E 1 (z) is the exponential integral:
E 1 (z) -0.57721 - tnz - t (9)
n1 nn!
5.3 Estimating the Mean Shipment Distance
In order to compute the estimates of the incident probabilities
for each container class, it was necessary to obtain an estimate of
the mean shipping distance of all hazardous material shipments
using that container class during the analysis year (1981). This
information is not directly avaitab e in the HAZMAT incident file
because the file contains information only for those shipments which
were involved in incidents.
-------
66
Below, we illustrate the derivation of an estimator for Let
X be a binary random variable indicating the occurrence (X1) or
non-occurrence (X0) of an incident. Given a shipment of distance
P (X=1 d) = d, and
P (X=0 d) = 1 - d
where is the combined incident rate (summed over all failure modes,
but unique for each container class).
Also, let f(d) be the overall shir r ng length distribution. The
conditional distribution (given an incident) is:
d f(d) d f(d)
f(dIXl)
I d f(d) E(d)
The first and second moments of this conditional distribution are:
E(d 2 )
/d f(dlXzl) (10)
E(d)
E(d 3 )
Id 2 f(dJXl) = (11)
E(d)
If f(d) is assumed to follow a Gamma distribution ; ith parameters u
and B, the first three moments are:
E(d) = a B
E(d 2 ) = aB 2 2 B 2
E(d 3 ) (a 1)(a2)
Thus, equations 10 and 11 become:
-------
67
E(dIX l) 8 (1 + a) (12)
E(d 2 IX:1) = (a 1)(a2) (13)
The parameters a and can then be determined from equations 12
and 13 by using the values of the conditional moments of the
shipping distances as computed from the data in the HAZMAT file.
The estimate of is then given by d = a8. Table 5.2 summarizes
the computed estimates of the mean shipping distances for seven of
the container classes analyzed in this study.
5.4 Fraction Release Model
The fraction release model is comprised of two sub-models: one
for the fraction of containers failed given an incident (the failure
model) and the other for the fraction spilled given a failure (the spill
model). We assumed that the fraction failed and fraction spilled
variables are dependent on both the container type and failure mode.
Using this assumption, we constructed linear models as follows:
F = aO a 1 X 1 ... + + + (14)
P = T 0 ... + 6 1 Y 1 + ... + 6 22 V 22 (15)
where F and P denote the fraction failed and fraction spilled, and the
Xs and Vs are binary variables denoting the container classes and
failure modes, respectively. For example, an observation
corresponding to container class 1 and failure mode 6 would have
X 1 1 arid V 6 1; the remaining independent variables would be zero.
-------
68
Table 5.2 Distance Distribution Summaries
Container N E(dIX 1) Var(dIXl) d
Class
1 60 790.38 596.55 0.7852 442.74 347.63
2 98 770.59 589.58 0.7259 446.48 324.10
3 99 942.45 651.88 1.1115 446.35 496.11
4 76 933.50 758.61 .5344 608.37 325.11
5 79 619.20 565.73 U.e .i 512.16 107.04
6 63 282.19 240.21 0.4022 201.24 80.94
7 103 858.68 637.48 0.8321 468.67 390.00
-------
69
The full regression models contain 29 binary variables which
define 8 container classes 7 and 23 failure codes, assuming that the
interaction terms are not significant in the analysis. The regression
coefficients in the models can be estimated using the spill data in the
HAZMAT file. Tables 5. and 5.4 summarize the estimates of the
dependent variables (failure and spill). Table 5.5 displays the
analysis of variance (ANOVA) for the full model regressions.
We then proceeded to test the hypotheses that the container
classes and failure modes are significant factors affecting the fraction
failed and fraction spilled. To test +c hypothesis on the fraction
failed, we constructed the following reduced models:
I I I
+ v v
O 11 2222
P = 0 + ... +
To test the hypothesis on the fraction spilled, the reduced models
became:
F = a 0 a:lxl . + a 7 X 7
p= 0 1 x 1 ... r 7 X 7
Tables 5.6 and 5.7 show the ANOVA results for the two reduced
models. Table 5.8 summarizes the significance tests for the container
class and failure mode effects. The results demonstrate the
7 These models did not include container class 8.
The tables include estimates from an independent regression for
container class 8 (open metal containers). They do not include the
estimates for container class 9 (other).
-------
Jo
FAILURE
M OIW
TABLE 5.3 Predicted Values of Fraction Failed by
Container Class and Failure Mode
CONTAINER CLASS
1 2 3 4 5 6
7 8
1
0.2968
0.2184
0.3290
0.2376
0.3218
0.7970
0.2736
0.1000
2
0.3059
0.2276
0.3382
0.2469
0.3310
0.8062
0.2828
0.2270
3
0.2555
0.1773
0.2879
0.1965
0.2806
0.7558
0.2325
0.2060
4
-
0.2408
0.1494
0.2336
-
0.1854
0.1250
5
-
0.1368
-
0.1560
0.2401
0.7153
0.1919
0.0400
6
-
0.2753
0.3859
0.2946
0.318i
0.8539
0.3305
0.1000
7
0.3549
0.2767
-
0.2959
0.3800
0.8552
0.3319
0.1000
8
0.2878
0.2095
0.3201
0.2288
0.3129
0.7881
0.2647
0.2960
9
0.2928
0.2146
-
0.2338
0.3179
0.7931
0.2698
0.1570
10
0.3586
0.2804
0.3910
0.2996
0.3837
0.8589
0.3356
0.2160
11
0.3884
0.3102
0.4208
0.3294
0.4135
0.8887
0.3654
0.2820
12
-
0.2894
0.4000
0.3087
0.3928
0.8680
0.3446
0.3630
13
0.3604
0.2821
0.3827
0.3014
0.3855
0.86 (17
0.3373
0.2730
14
0.2894
0.2112
0.3218
0.2304
0.3146
0.7898
0.2664
0.1870
15
0.2848
0.2065
-
0.3099
0.7851
0.2617
0.1550
16
-
0.2010
-
0.2202
0.3043
-
0.2562
0.2870
17
0.3848
0.3066
0.4172
0.3258
0.4099
0.8851
0.3618
0.1430
18
-
-
-
-
-
0.9825
19
0.4440
0.3658
0.4764
0.3850
0.4691
0.9443
0.4210
0.0540
20
-
Ψt774
0.2880
0.1966
0.2807
0.7559
0.2326
0.1930
21
-
0.2805
0.3911
0.2997
0.3839
0.8591
0.3357
0.2650
22
0.2768
0.1985
0.3091
0.2178
0.3019
0.7771
0.2537
0.2540
27
-
0.6768
0.7874
0.7801
1.0000
0.7319
0.1000
-------
71
TABLE 5.4 Predicted Values Of Fraction Spilled by
Cor tai ier Class and Failure Mode
FAILURE
MODE CONTAINER CLASS
1 2 3 4 5 6 7 8
1 0.4471 0.5622 0.831S 0.5843 0.3908 0.2399 0.3621 0. 1000
2 0.3815 0.4966 0.7659 0.5187 0.3253 0.1743 0.2965 0.4170
3 0.4137 0.5 κ 0.7982 0.5510 0.3575 0.2065 0.3287 0.3510
4 - 0.7189 0.4717 0.2782 - 0.2495 0.5000
5 - 0.4767 - 0.4989 0.3054 0.1544 0.2766 0.5500
6 - 0.4033 0.8726 0.4254 O.i . ., 0.0810 0.2032 0.1000
7 0.4586 0.5737 - 0.5959 0.4024 0.2514 0.3736 0.1000
8 0.2718 0.3869 0.8562 0.4090 0.2155 0.0646 0.1868 0.1980
9 0.2075 0.3226 - 0.3447 0.1512 0.0003 0.12 5 0.0960
10 0.1691 0.2842 0.5535 0.3063 0.1129 0.0003 0.0841 0.1770
11 0.2120 0.3271 0.5964 0.3492 0.1558 0.0048 O. 270 0.2200
12 - 0.4200 0.6893 0.4421 0.2487 0.0977 0.2199 0.3740
$3 0.2936 0.4087 0.6780 0.4308 0.2374 0.0864 0.2086 0.4220
14 0.2789 0.3940 0.6633 0.4161 0.2227 0.0717 0.1939 0.3310
15 0.2366 0.3517 - - 0.1804 0.0294 0.1516 0.2120
16 - 0.3622 - 0.3843 0.1909 - 0.1621 0.1540
17 0.3069 0.4220 0.6913 0.4441 0.2506 0.0997 0.2219 0.4470
18 - - - - 0.0075 -
19 0.2716 0.3867 0.6560 0.4089 0.2154 0.0644 0.1866 0.6670
20 - 0.6703 0.9397 0.6925 0.4990 0.3480 0.4703 0.8580
21 - 0.5332 0.8025 0,5553 0.3619 0.2109 0.3331 0.3310
22 0.4306 0.5457 0.8150 0.5678 0.3744 0.2234 0.3456 0.4180
27 - 0.4538 0.7232 - 0.2825 0.1315 0.2537 0.IU0U
-------
72
Table 5.5 ANOVA Table for Full Model
(a) Fraction failed
SOURCE DF SS MS
Regression 30 1457.13 48.57
Residual 7774 762.87 0.10
Total 7804 2220.0
(b) Fraction spilled
SOURCE DF SS MS
Regression 30 961.78 32.06
Residual 7774 828.22 0.11
Total 7804 1790.00
DF degrees of freedom
SS = sum of squares
MS mean square
-------
73
Table 5.6 ANOVA table for Reduced Model Testing for Container
Class SignifIcance
(a) Fraction failed
SOURCE SS MS
Regression 23 1293.90 56.26
Residual 7781 926.08 0.11
Total 78ti4 2220.0
(b) Fraction spilled
SOURCE DF SS MS
Regression 23 841 .03 32.56
Residual 7781 948.97 0.12
Total 7804 1790.00
OF degrees of freedom
SS sum of squares
MS mean square
-------
74
Table 5.7 ANOVA Table for Reduced Model Testing for Failure Mode
Significance
(a) Fraction failed
SOURCE DF SS MS
Regression 8 1436.74 179.59
Residual 7796 783.26 0.10
Total 7804 2220.0
(b) Fraction spilled
SOURCE DF SS MS
Regression 8 908.01 113.50
Residual 7796 881.99 0.11
Total 7804 1790.00
OF degrees of freedom
SS sum of squares
MS mean square
-------
75
Table 5.8 F-test Summaries
FACTOR SUBMODEL COMPUTED SIGNIFICANCE
F
Container Fiaction Failed 237.54 p < 0.01
Class acti Spilled 161.92 p < 0.01
Failure Mode Fraction Failed 9.44 p < 0.01
Fraction Spilled 22.94 p < 0.01
-------
76
significance of both effects at the 1 percent level.
5.5 Fraction Release Estimators
Let and R 1 denote the random variables fraction failed,
fraction spilled and fraction released for failure mode ii, with means
Vfji u and rj respectively. Thus:
R 1 = F P 1
Assuming that and are independent:
rj = Ufj Wpj
Using r 1 to denote the estimate of UrjP we obtained:
= f p
where f 1 and pj are the mean response estimates obtained from the
models in equations 14 and 15.
Recall that and denote the probabilities of incidents
occurring by failure mode j enroute and at shipping terminal points,
and that and 8 are their estimators. Let ur and rt denote the
mean fraction released per mile shipped and at terminal points,
respectively. Let r and rt denote their respective estimators. Then:
23
r = r.A. r 1 X (16)
j2
rt £ r.8. (17)
j ii
-------
77
where A, corresponding to the failure mode releasing vehicular
accident, i considered an input variable which need not be
equiva(ent to th overall mean truck accident rate ( ) used in
estimating thη o% r ipcid nt probabilities, and O . !n fact,
depending on roadway tc., various values of A cap be used in
computing the releqe fr c ipp ip equation 16.
5.6 Fraction ReJc r tirngtes
In the previpqs se4jFm we derived several estimators which
are required to estimate the expected fraction releas?d. We computed
estimates for the expected fraction failed and fraction spilled as
shown in Table 5.3 and 5.4. In addition, we computed estimates for
the mean shipment distances for each container class (see Table 5.2).
Finally, we require an estipiate, A (and A), of the releasing truck
accident rate. In Chapter 6 we will discuss the determination of
estimates for the truck accident rate. In computing A (and A).
however, we must account for the fact that not all truck accidents
result in a release. We derived an estimate of 0.2 for the fraction of
truck accidents in which a spill occurs. This was based on the
following factors. First, the 1981 FRA Accident/Incident Bulletin (15)
indicates that in 601 train accidents consisting of 2,770 cars carrying
hazardous materials, 109 cars released. Second, previous work by
Geffen [ 10] indicates that tank trucks involved in accidents are
approximately 10 times more likely to spill than rail tank cars. These
two factors yield an estimate of 0.4 which we adjusted downward to
-------
78
compensate for the fact that the damage threshold for an FRA
reportable accident is higher than the threshold used in the HAZMAT
file.
Table 5.9 summarizes the estimates of the expected fraction
released both enroute and at terminal points for the container classes
considered in this analysis. Note that the expected fraction released
per mile shipped is expressed in terms of X, a releasing accident
rate which may vary depending on transport link characteristics.
Estimates for X are, obtained by multiplying the accident rates given
in Chapter 6 for various roadway types and traffic volumes by 0.2.
The aggregate accident involvement rates (releasing accidents per
million truck miles) are summarized for different highway types below:
Interstate 0.13
U.S. and State 0.45
Urban 0.73
Composite 0.28
In order to evaluate our results, we compared the estimates for
tanks in Table 5.9 with the results of the Bercha study [ 9] for tank
trucks and vacuum trucks, and the PNL studies [ 8,10] for tank and
tank-trailer combination trucks. The PNL studies report incident
probabilities in a 210 km shipment of 3.68 x 10 and 3.57 x 10 for
propane and gasoline carrying trucks, respectively. These values
translate to an incident probability per mile of 2.8 x 1O 7 which
compares favorably with our estimate for the fraction released per
-------
78
Table 5.9 Estimates of Fraction Released by Container Class
Container Expected Fraction Expected Fraction
Class Released Per Released at Terminal
Mile Shipped Points
1 1.3 x io6 + (.13 A) 1.4 x
2 2.6 x io_6 +(.12 A) 4.0 x 10
3 1.7 x io 6 +(.27 A) 2.6 x 1O
4 4.3. x 1o6 +(.14 A) 5.2 x
5 1.3 x io6 4(12 A) 6.1 x 10
6 4.2 x io.8 +(.19 A) 7.6 x io6
7 2.4 x 106 +(io A) 2.9 x l0
8* 7.5 x io 1.2 x
*estimate associated with the release fraction during accident is not
reliable.
-------
79
mile of 1 x 1O . The Bercha stud reports release fractions per
mile of 2.02 x and 1.68 x 10 for vacuum trucks and tank
trucks, respectively. In addition, Bercha reports fraction release
estimates during loading/unloading of 4.6 x 1O and 2.4 x 1O for
vacuum trucks and tank trucks, respectively. Our results for
incidents enroute are in general agreement with Berchas. For
incidents at terminal points, however, our results are two orders of
magnitude lower. This apparent discrepancy could result from
under-reporting of HAZMAT small spill incidents at terminals. If we
.
remove the very small spills from the 8 rcha analysis, the resulting
release fractions during loading/unloading for both vacuum and tank
trucks become 2.4 x 10 . These are still three times higher than
our estimate of 7.6 x io .6.
5.7 Errors of the Estimates
There are several sources of error which affect the reease
estimates in Table 5.9. These can be categorized as modeling errors
and estimation errors. In this section, we are interested only in the
estimation errors and their implications.
Recall tha in equations 4 and 5, there are three factors to be
estimated: A, the releasing truck accident rate; d the mean
shipping distance for the container class; and the incident frequency
ratios. In view of the functional form of the estimators, the errors
in the aforementioned factors are multiplicative. That is, a 10% error
-------
80
in A and a 10% error in (n +1)/n 1 yields a 21% error in X . The
error in A, in turn, is multiplicative in the errors in the accident
rate estimates and the estimates of the fraction of accidents which
release. In order to gauge the total error, we looked at each of the
factors individually.
The frequency ratios which we derived from the HAZMAT data
could be affected by under-reporting of incidents. There is strong
evidence to suggest that this occurs. However, if the
under-reporting is uniform across all failure modes, our estimates are
not affected. It is our view that accidents are not as likely to go
unreported are other incidents (particularly at terminals) and this
would lower our estimates.
The estimates of the truck accident rates derived in this study
are within the range of previously reported findings. As an average
of rates representing varied highway and traffic volume conditions,
the composite rate used in our analysis is lower than what was used
in the PNL (8,10] and Bercha (9] studies. This again would tend to
lower our estimates.
With regard to the estimate of the fraction of accidents which
release, it may be argued that our estimate of 0.2 is high. For
example, it has been suggested that one can use the fatality rate as
a proxy for the releasing accident rate. From data reported in
NHTSA [ 26], 8.6% of single vehicle truck accidents result in a
fatality. NHTSA also reports injury rates of 24%. Thus, a factor in
-------
81
the range of 0.08 to 0.24 appears reasonable.
There are other factors whose errors affect the computations of
the final fraction release estimates. These include sampling errors in
the estimates of the fraction spilled given an accident, and errors in
the estimation of the shipping distances by container types. The
magnitude of these errors is given by the standard error of the
estimates and is less than 20%.
As an illustration of the overall error effects, consider the
possibility that we underestimated the accident rate by 25%,
overestimated the fraction of accidents by 100%.
overestimated the shipping distance by 20% and underestimated the
frequency ratio at terminals by 20%. For the above situation, the net
error in the incident probability estimates would be approximately 44%.
5.8 Results and Implications
Using the HAZMAT data, we estimated the fraction of containers
failed and the fraction spilled for each defined container class and by
each failure mode. We also computed the probabilities of incidents
occurring in two categories: enroute and at shipping terminal points.
These estimates enabled us to determine the overall fraction released.
The results of our analyses indicate that in terms of their
order of magnitude, the expected fractions released per mile shipped
range from 1c18 to depending on the container class. The
expected fractions released at terminal points range from i0 6 to
depending on the container class.
-------
82
Our computed estimates indicate that:
(1) The release rates for tanker trucks are much lower than
for other container types.
(2) The expected amount released at terminal points is one to
three orders of magnitude higher than the amount
released en route.
(3) The expected release fractions during transport are
poteritiafly as high as the release fractions at disposal
sites and treatment facilities which range from 1O to
1O for routine spillage and 1O 5 to 1O for accidental
spillage [ 16].
-------
83
CHAPTER 6
ESTIMATING THE TRUCK ACCIDENT RATE
After we derived the expected fraction release estimates per
mile shipped in terms of the truck accident rate (Chapter 5), we
performed an analysis of the truck accident rate data (see Chapter 3)
to derive estimated accident rates for different roadway types. We
defined the truck accident rate as follows:
N xl o6
y=
TADTx Lx365
where:
y Is the accident rate (accidents per million truck
miles).
N is the frequency of truck accidents for the analysis
year.
TADT is the average daily truck volume.
L is the length of the section over which the volume
and accident data were collected.
Although the truck accident rate for a given section of road is a
function of many traffic and driver related factors, the primary
interest for the present analysis is in the dependence of the accident
involvement rates on different highway types, and traffic and truck
volume levels.
Previous research in this area includes the work of Vallette, et
al. (17], ADL [ 18], FHWA (19], BMCS [ 20], Zeiszler (21], Scott and
ODay [ 22], Voo [ 23], Smith and Wilmot (24], Meyers [ 25] and others
(see NHTSA (26]). In several of the above studies, accurate truck
-------
84
exposure data was not available. In others, only one highway type
was considered. The Vallette study provides reasonably accurate
estimates for accident rates ranging from 0.43 to 5.24 per mtllion
truck miles for different truck and highway types. However, traffic
volume levels are not considered.
6.1 Analysis
The truck accident and volume data collected from California,
Texas and New Jersey (see Chapter o) included a wide range of
traffic and truck volumes, and four distinct highway types. From
this 3-state database, we obtained data on the volumes and
frequencies of accidents for trucks of 2-axle dual tires and larger.
We used this subset because it is most representative of the vehicles
used to transport hazardous materials.
To test the statistical significance of any differences in accident
rates for different highway and traffic volume levels, we conducted
an analysis of variance (ANOVA). The analysis of the data from
California and Texas was conducted as a fixed effect, three-factor
(truck percentage, traffic volume and highway type), mixed design of
unequal sample size. We nested the traffic volume factor (ADT)
within the highway type factor because the California data seem to
correspond to much higher ADT volumes than did the Texas data.
Table 6.1 shows the means and standard deviations for each cell (SR
= state highway, U.S. = U.S. highway and lH interstate highway),
Excluding 2-axle pick-up and panel trucks.
-------
85
Table 6.1 Cell Statistics for California and Texas
Highway %Truck ADT(x10 3 ) N Mean St.dev
Type
SH <7 0-25 15 5.623 6.456
25-SO 7 1.389 0.675
>50 4 1.586 2.040
>7 0-25 23 1.014 1.034
25-50 4 0.883 0.793
>50 4 0.554 0.317
US <7 0-25 5 7.563 9.379
25-50 6 2.065 3.592
>50 6 1.590 1.544
>7 0-25 11 1.219 0.828
25- 0 5 0.536 0.337
>50 1 0.600 0.000
IH <7 0-40 2 0.425 0.352
40-80 3 1.469 1.617
>80 11 0.951 0.549
>7 0-40 27 0.413 0.386
40-80 11 0.624 0.435
>80 4 0.733 0.466
-------
86
and Table 6.2 shows the group statistics. The analy i.s
in Table 6.3 demonstrates the significance of the main effe.
percentage and ADT at the 5 percent level.
We conducted our analysis of the New Jersey dut .
three-factor (truck percentage, traffic volume and nunth.r c f
intersections) crossed design. Our analysis of the New .Jer /
also indicates the significance of the main effects at the 5
level. The results are summarized in T. ..3les 6 i. 6.5 and 6.C.
6.2 Results and Implications
The analysis of the truck accident rate data yielded iI-
following estimate for the accident involvement rates (acciden .
million truck miles):
lnterstates 0.65
U.S. and State Highways 2.26
Highways with interrupted 3.65
flow due to intersections
These results fall within the range of previoti Jy -
estimates and demonstrate the difference in the accident -
various highway types. Furthermore, the analysis in the p .
sections shows that the truck accident rate is dependent c. tb
total traffic volume and the percentage of trucks in th rm
stream. These results suggest that in applying the i .. .
provided, cell means should be used in lieu of aggregate meals ;i
-------
87
Table 6.2 Group Statistics for California and Texas
Factor Level Count Mean St. dev.
Highway SH 57 2.271 3.910
US 2.248 4.295
IH 58 0.632 0.584
Truck % <7 59 2.981 4.829
>7 0.740 0.728
ADT 0-25 72 2.298 4.427
25-50 38 1.032 1.550
>50 39 0.973 0.974
-------
88
TabI. 6.3 ANOVA Table for New Jersey
Sourcs of Degrees of Sum of Mean F Level of
Variation Freedom Squares Square Statistic Significance
Main
Factors
Highway 2 31.81 15.90 1.91 0.153
Type(H)
%Truck(T) 1 66.99 66.99 8.03 0.005
Nested
Factor
ADT 6 133.23 22.20 2.66 0.018
within H
Interaction
T and H 2 24.25 12.12 1.45 0.237
T and 6 91.37 15.22 1.83 0.099
ADT
within H
Error 131 1033.85 8.33
-------
89
Table 64 cell Statistics for New Jersey
Number tTruck ADT(X10 3 ) N Mean St. dev
Int.rnc. (veh/day)
per Smiles
0-8 c7 0-20 10 4.709 2.489
20-40 3 2.878 1.560
>40 3 1.391 0.283
>7 0-20 8 1.875 0.842
20-40 5 1.262 1.531
40 2 0.457 0.034
>8 <7 0-20 2 10.28 0.022
2040 7 6.633 2.747
>40 1 3.454 0.0 ,00
0-20 3 4.571 3.598
20-40 6 2.969 0.896
>40 4 2.406 1.176
-------
90
Tabs. 6.5 Group Statfatics for N.w .Mrs.y
Factor Lavil Count M.an St.d.v.
Interiec. 0 -8 29 2.703 2.286
(number/5miIes) . $ 23 4.852 3.074
%Truck 7 26 2.293 1.771
ACT 0-20 21 4.410 3.112
20-40 21 3.771 2.816
>40 10 1.817 1.184
-------
91
Table 6.6 ANOVA Table for New Jersey
Source of Degrees of Sum of Mean F Lavs of
Variation Freedom Squares Square Statistic Significance
Main
Factors
%TRUCK(T) 1 62.06 62.06 16.22 0.000
ADT 2 64.19 32.39 8.57 0.000
lntersections(I) 1 78.19 78.19 20.44 0.000
Interaction
T and ADT 2 14.17 7.08 1:85 0.170
T and 1 6.31 6.31 1.65 0.206
ADT and I 2 6.84 3.42 0.89 0.417
T aM ADT 2 2.41 1.20 0.32
and I
Error 40 .153.02 3.82
-------
92
sufficient information is available to Identify the highway type and the
traffic volumes. Furthqrniore, if in a given situation one has
available more accurate accident rate data, then the data should I e
used in lieu of the rates provided In this report.
For the purpose of computing the fraction release estimates in
equations 16 and 17, we derived a composite truck accident rate of
1 .4 accidents per million truck miles based on a weighted average of
the rates previously mentioned.
-------
93
CHAPTER 7
ESTiMATING THE COST OF
TRANSPORTING WASTE 10
This chapter describes how we estimated the cost of
transporting hazardous wastes by truck Briefly, our procedure was
as follows. First, we reviewed the existing literature directed at
estimating the cost of transporting hazardous wastes. From our
review, we identified seven studies that addressed the issue of
estimating the cost of transporting hazardous waste by truck. All of
these studies considered this issue within the larger framework of the
total cost and risk of hazardous waste treatment at a regional level.
Next, we selected the mo*t comprehensive of these methodologies
and developed a revised cost procedure using some of its assumptions
and modifying others. Finally, we determined the accuracy of our
costing procedure by comparing its estimated results with the actual
rates charged by haulers.
7.1 Literature Review
In a report to the Environmental Council of Alberta concerning
the transp3rtation risks involved in treating hazardous waste
substances, Bercha and Associates (9) addressed the costs of
transporting hazardous waste by segmenting costs according to trip
length:
10 The RCRA Risk/Cost Analysis Model uses these costing assumptions
and unit costs, but uses a different accounting procedure.
-------
94
Trip Length Cost ( Canadian !I Cost ( U.S . !1.
Per Tonne-Kilometer f ! Ton-Mile
0-100 Km (62 mi) 0.120 0.176
> 100 Km (62 ml) 0.080 0.fl7
The Bertha analysis did not differentiate its calculated costs by
truck capacity or material transported. Also, we had to make
assumptions about two items that were not reported in the Bertha
paper. First, we assumed that trip length corresponds to the
one-way trip dietance but that the costs of deadheading back to the
point of origin are embedded in Berthas cost estimates. Second, we
assumed that trip length was segmented to reflect the decrease in per
ton-mile costs that will occur with longer trips (fixed costs are
distributed over a larger base).
A study by Boor, Alien and Hamilton (27] addressed
transportation costs as part of an assessment of hazardous waste
generation and treatenvent capacity. Boor-Alien assumed that all
hazardous waste would be transported by either 6,000 gallon tank
trucks or flatbed trucks carrying 80 drums. Their report implies
that trucks would be traveling at full capacity. On the basis of
interviews with facility operators, Boor-Allen posited three different
-------
95
rules of thumb for truck transport costs
Method Cost j J
Flat rate per hour $30 - $40
Flat rate par mile, round trip $1.50 - $3.00
Fixed costs plus variable cost $100 - $150 minimum charge and
(usually applied to shorter trips) $1.00 to $1.50 per mile
It should be noted that Booz-Allan did qualify its work by stating
that not all facility operators use these rules of thumb.
The Booz-Allen study does not indicate the conditions under
which each costing method is most appropriate. The study also
assumed that the costs for transporting waste by tank or drum are
similar, and it did not recognize the expected decrease in per-mile
costs associated with longer trips. Finally, the assumption that
trucks travel at full capacity is not supported by analyses which
have been conducted on hazardous waste shipment characteristics
reported in Chapter 4. Consequently, the estimated costs are likely
to be biased on the low side.
In its study of the New York State hazardous waste management
program, Camp, Dresser and McKee (CDM) conducted telephone
interviews with haulers operating within the state (281. CDM
obtained estimates for a 75 mile one-way trip using 4,000 gallon tank
-------
96
trucks. Their cost estimates (including all fees, tolls, gas and
wages) ranged from $1.14 to $4.80 per truck-mile depending on
distance, waste type and quantity. For their purposes, Camp,
Dresser and McKee usad an average cost of $1 .25 - $1 .50 per mile.
The importance of this study is not in the assumptions CDM
adopted (which suffer from the deficiencies described previously In
the Bercha and Booz-Aflen discussions) but in the information
obtained in conversing directly with operators. The operators
themselves identified trip distance, at -- - t size and waste type as
being important factors in determining truck transportation costs.
Transport cost was treated quite generally in a study of
hazardous waste management in Massachusetts [ 5]. The Massachusetts
Bureau of Solid Waste Disposal assumed that waste would be
transported in either 80 drum trucks or 4,400 gallon tanker trucks,
and that trucks only travel, at full capacity. Costs were estimated at
$1.00 $3.00 per truck-mile (one-way trip), which is equivalent to
$0.06 - $0.18 per ton-mile. The Massachusetts study adopted a rate
of $0.12 per ton-mile. No additional insights could be gained from
reviewing this costing approach. Beyond assuming that shipments are
only made at full capacity, the methodology suffers from assuming
that per-mile costs remain constant, irrespective of trip length and
material transported.
In contrast to the variable cost structure established in the
first four studies, Arthur 0. Little (ADL) developed a more
sophisticated approach for its assessment of hazardous waste
-------
97
management facilities in New England [ 29]. ADL. recognized that the
real cost of transporting wastes consists of a fixed cost (capital
amortization, insurance taxes, salaries, fringes, supervision, general
and administrative) which is independent of the shipping activity and
a variable cost (fuel, tires, lubrication, maintenance) which is likely
to be a function of trip distance.
In developing its cost formulas, ADL assumed that a truck is in
service 2.000 hours a year and, during the time that the truck is in
service and on the road, the average Lravel speed is 40 mph. ADL
further assumed that the truck operates at capacity when a shipment
is made and returns empty to the point of origin. Using these and
other assumptions (see Table 7.1), ADL conducted its analysis for
6,000 gallon tank trucks and stake trucks capable of carrying thirty
55-gallon drums.
Using this information, ADL derived the following cost
functions:
Tanker CT = 0.084 2.45/d
Stake truck CT = 0.237 + 1l.01/d
where:
CT = cost in S/ton-mile
d = one-way trip distance (miles)
The major advantages of ADLs approach are: 1) its detailed
transportation cost components, 2) its recognition that some costs are
fixed while others are variable, 3) its use of different truck types
and 4) its use of unit costs which decrease as a function of trip
-------
98
Table 7 1
AOL Cost Assumptions - Now England
ma Ty :
Cost
LosdeieW tMlaalnq PuS
Ph.d Co . . (8 /yrl
apual amantanon
8yn 024%-0.292
salaries & fringes
.11 t7SIhr
wpenuion (40% of atnel
maurance and taxes
G&A i0%
8000 Gallon Tanler
Load a w a tans
aoao
2 hours
18.060
25 .500
10.200
4.000
66.760
51576
81,338
S ksTmdc -20ftbs4
ses on dru m .e7
524.000
3hgsgrs
7.008
25 1 500
10200
- 4,000
46.708
4,671
51,379
Opsradnq corn lWm*)
Fuul Snggl ΆOOllΨlonl
Tires and lubrication
G&a*S10%
0.17
0.08
22
0.20
2 2
0.29
0.11
0.03
0.14
0.19
Source: Arthur D. Little, Inc. , Plan La Develoffi.ent Eazardous Waste
!3 !.2i Facilities in the Rev Region, Volume I : Arn,s d4 tea .
prepared for the 14ev Eng land Ragianni Co 4 aion, September 1979.
(Ompg@ IOOWgsIIon )
-------
99
distance.
The drawbacks of this work are:
(1) The estimates of capital and operating costs were not
validated against actual records.
(2) It was assumed that trucks operate at full capacity
during transport.
(3) It was assumed that trucks are constantly in demand and
available for service.
These assumptions contribute a bias towaru Lnderestimating the real
transport cost per shipment.
ADL revised its 1979 costing procedure for a study of
hazardous waste quantities and facility needs in Maryland (303. The
primary modifications were:
(1) Trucks were assumed to be in service 80 percent of the
time.
(2) A line item for profit (5 percent of non-capital related
expenses plus general and administrative expenses) was
included.
(3) A roll-off container truck with capacity for eighty
55-gallon drums was included.
(4) The component costs were updated to account for
inflation and other changing market conditions. For the
Maryland study, ADL contacted operators and
manufacturers in the U.S. to verify the plausibility of
its component cost assumptions.
-------
100
ADL.s estimates of the cost per ton for one-way trip distances
of 50 and 100 miles for tank trailers and stake trucks transporting
roll-off containers appear in Tables 7.2 and 7.3. In their report,
ADL described the following generalized cost formulas:
Tanker (25 tons) CT = 3.09 0.115 d (s/ton)
Stake truck (18 tons) CT = 11.66 0.312 d (s/ton)
However, we applied these formulas w m c information in Tables 7.2
and 7.3, and obtained quite different results between the formula end
table:
Distance Truck Type Cost/Ton Cost/Ton
Estimate EstImate
in Table by Formula
50 miles Tank $7.91 $8.84
100 miles Tank $13.22 $14.59
50 miles Stake $12.49 $27.26
100 muGs Stake $19.71 $42.86
These discrepancies, particularly for the stake truck, raise serious
questions about the validity of the Mary lnd cost formulas. However 1
the basis for the cost estimates in Tables 7.2 and 7.3 appear to be
sound.
%flh nil
-------
iU. .
Table 7.2
AOL Pricing Procedure - Tank Trailer (Bulk Liquid -
25 tons
Typical Trip
One way distance
Tonnage per trip
Loading/unloading time
Tins on raid
Total trip c ms
Capital Coat (1978 3 )
Power unit
Tank trailer
ical Related Rourly targ.a
tntsrsse at 132
Deprsc lat ion
$auCa itaL Related Rourly Q srgss
Driver s salary
Suprv is ion
Insurance
License & tan
Pu fliLe arges
Fuel and oil
Tires, nsinc.uancs and repair
4.3 hra.
5.4 bra.
$103.14
32.00
5135.1.4
13.51
$148.65
7.43
$156.08
41. . 72
$197 10
$7.91
9 bra.
8.4 bra.
$180.44
64.00
$224.44
22.44
12.34
$259.22
71.32
$13.22
Source: Arthur D. Little, Inc. Eazardous Waste Quantities Facilit 1eeds
4aryLand . prepared for Razardous tJasta Facilities Siting Board and
Iaryla d Envtronn ntal Science, August 1981. -
50 miles
100 miles
25 tons
25 tons
2 bra.
2 bra.
2.5 bra.
5 bra.
4.3 hrs.
7 bra.
$40,230
$40,230
26.000
24.000
$64,250
$64,250
$4.82
$4.82
2.36
2.86
57.b8
7.6.8
512:50
%:
2.10
2.10
2.00
2.00
$19.10
$19.10
$0.20
$0.20
0.12
0.12
$0.32
$0.32
Total trip time
aarg.ab1a trip tins
(1.2 x total trip tins)
4on-capical related hourly costs
Per nil. charges
G & A @ 102
Profit @ 5%
Capital related hourly coats
Cost per con
-------
102
Table 7.3 ADL Pricing Procedure - Stake Truck (Drummed Liquid,
Solid; Bulk Liquid - 18 Tons)
typ .ca1 . Trip
One ay distance
tonnage per trip
i.oading/ualotdiztg time
tins on road
Total trip time
Csptal Cost (1978 $ )
Povec un .t
ri1efr
Rolloff cont*uier _______
Canital Related Bourlv ar,es
Interest 152
Dpr.ciatioe _____ _____
Related ur1i
Driver S
Supervis ion
Insurance
License and tans. ______
Per Mile Qiart. .
Fuel and oil
Tires, maintenanc, and repair _____ _____
5.3 bra. 8 bra.
6.6 hr .. 9.6 bra.
$126.06 $183.26
_ 12.00 64.00
$L58. 06 $24T.36
L3.81. 24.74
$173.87 $272.10
8.69 13.61
$182.56 $285.71
42.31 69.21
$224.87 $334.83
$12.49 $19.71
Source: Arthur D. LittLe, Inc. Razardous Waste Quantities Meets
in Maryland . prepared for Eazardous Vast. Facilities Siting 3oard and
Maryland Environm tal Sciante, August 1981.
50 miles
100 ails.
18 tOn.
18 tom.
3 bra.
3 hrs.
2.3 hr..
5 bra.
5.3 bra.
8 hra.
$40,250
$40,250
14.300
14,300
2.800
S 7 ,550
2 800
T!? !
54.32
$4.32
2.09
2.38
$6.41
$7.20
2.50
$12.50
2.50
2.50
2.10
2.10
2.00
TIT1T
2.00
$1 9.10
$0.10
$0.20
0.12
0.12
Transport Cata
Total trip t me
Chargeabl. trip time
(1.2 x total trip time)
Moncapital rslac.d hourly cost.
Per i1s charge.
C 6 A 102
Profit @ 52
CapitaL zela:ed hourly costs
Cost per con
-------
1V1
.1.
ADLs overall approach corrects for many of the first five
studies methodological problems. The major remaining problems are:
1) ADL assumed thit trucks are fully loaded and 2) although it
consulted operators on th. component cost estimates ADL did not
examine actual cost records to determine if its total costs were
representative of actual costs.
For an earlier version of the RCRA Risk/Cost Analysis Model,
ICF examined the costs of transporting waste by 6,000 gallon tank
trucks for one-way trip distances of nd 250 miles (16). ICF
assumed that on-site transportation costs were included in treatment
and disposal costs (this assumptiofl appears to be implied in the other
six studies)
ICF formulated a procedure similar to that developed by ADL.
However, unlike ADL ICF did not formulate the following cost
factors:
(1) Supervisory labor.
(2) Interest on capital.
(3) Insurance.
(4) Tax.
(5) General and administrative.
(6) Profit.
The ICF procedure suffers from the same deficiencies as ADLs
Maryland methodology and, in addition, is not as comprehensive. For
these reasons, the 1CF approach appears to be less suitable for
adoption than the ADL methodology.
-------
104
In ,immsry, the methodologies we reviewed fall into two major
categories: variable cost models and total (fixed plus variable) cost
models. The total cost models are more sophisticated in th&r
treatment of component costs; thus they are likely to b. more
representative of ths real cost of op.rating service. Of th. total
cost models. ADLi Maryland model appears to be th. most complet.
although som. deficiencies still remain.
Below, we describe a revised procedure that was dsv&op.d to
address thes deficiencies.
7.2 Revised Procedure
We devised a costing procedur. based on ADLs Maryland study
cost assumptions with the following modifications:
(I) We updated costs into 1983 terms using th. consumer
price in : e, wher. appropriate.
(2) We assumed average trip distances and shipment sizes
based on the results of the analysis of hazardous waste
shipment characteristics.
(3) We compared the revised cost formulas to actual price
quotes from waste haulers in order to establish the
accuracy of the revised procedure.
We estimated transport costs for 6,000 gallon tankers and 18-ton stake
trucks. As in the case of the ADL study, we segmented costs into
fixed and variable costs, as described in Table 7.4.
-------
105
Table 7.4 Coat .Aa. ptiona for Ravised Procedure
6000 Gallon Stake Truck
Tanker ( 18 Ton)
n.xth COSTS
Capital Coat $90,400 $81600
Capital Aaoritiution 18 170
8 yra. 9 12% 0.201 , 16,402
NonCapital Fixed Charies (3983*)
Drivers Salary: 14.64/hr z 2000 29,280 29,280
Supervision: 2.93/hr. x 2000 5,860 5,860
tusuranc*: 2.3Olhr. x 2000 4,200 4,200
License and Tax: 2.00/hr. x 2000 4,000 4,000
Total Capital and Fixed arg.. 61,510 59,742
C + A 9 102 o,i.31 5,974
Profit 9 32 3,383 3,286
TOTAL FIX COSTS/TI 71,044 69,002
VARIABLE COSTS (5/mile)
fuel and Oil $0.23 $0.23
Tiru, Main. and Repair 0.14 0.14
C + A 9 102 0.34 0.04
Profit 9 SR 0.02 0.02
TOTAL VARIABLE COST/MILE $0.43 $0.43
*User Consumer Price tt4ex (CPI) figures for urban wages, th. inflation
rate has been as follovs: 1981 10.4%, 1982 6.12.
-------
106
7.2.1 Averaqe Cost Approach 6.000 Gallon Tanker
Analysts often require average cost information in order to make
policy decisions where detailed information on shipment characteristics
is not available. This approach can be facilitated by assuming an
average shipment size and trip length for a typical shipment. Below,
we examine average costs for tanker transport, assuming that the
tanker is carrying liquid materials.
We asssumsd that: 1) the utilization rate is 80 percent (In
service 1,600 hours por year), 2) time . the road is based on an
average speed of 40 mph and 3) th. loading/unloading time is 2 hours
for ach shipment. Based on the analysis of hazardous waste
shipment characteristics, the weighted mean trip length is 84.2 miles
and th. average shipment size ii 3.171 gallons. equivalent to 13.21
tens. These inputs, coupled with the information in Table 7.4,
yielded the following results:
84.2x2 miles
vsrage time per shipment 2 Pin. 8.21 hrs.
40 mph
1600 his
avsrags trips p.r year = = 257.65
6.21 hrs
71,044
average fiied cost per trip = $275. 14
257.65
average variable cost per trip 0.43 x 84.2x2 $72.41
-------
107
average total cost per trip = 275.74 72.41 = $348.15
348.15
average cost per leaded mile $4. 14
84.2
1.14
a erags cost per loaded ton-mile - $0.31
13.21
Thus we determined that the average cost per loaded mUe of tanker
transport is $4.14 and the average cost er loaded ton-mile is $0.31.
7.2.2 Average Cost Approach - Stake Truck
We used th. same time, distance and quantity assumptions as in
the previous case, with the fcllowing exceptions:
(I) Loading/unloading time was assumed to be 3 hours.
(2) Average shipment size was assumed to be 11.63 tons.
Tb. analysis proceeded as follows:
84.2 c2 miles
average time per shipment 3 hrs. = 7.21 hrs.
40mph
1,600 hrs
average trips per year = 221.9
7.21 hrs
$69. 002
average fixed cost per trip = $310.96
221.9
average variable cost per trip = 0.43 x 84.2x2 = $72.41
average total cost per trip = 310.96 72.41 6383.37
-------
108
$383.37
average cost per ioaded mile : $455
84.2
$4.55
average cost p .r loaded ton-mile $0.39
11.83
Tb. average costs per loaded mu, and loaded ton-mil. are larg.r for
stak. trucks than tankers. This is due to th. smaller loads
associated with stak. trucks.
7.2.3 yin Cost Formulas
Whsn details on specific shipments ar. available, It 5 extremely
useful to hay, formulas which can be used to estimats th. cost of
transport. Below, we discuss how formulas wer. derived for tankers
and stak. trucks.
After defining F as annual fixed cost, )C as one-way shipment
length (mites), V as shipment size (tons) and Z as loading/unloading
time (lws), we expressed the average cost per loaded mile as:
F 1
clm(S/loaded mile) (0.43x2)
(1600/(.05XZ)) 2X
For tankrs: F = $71 ,044 and Z 2. Therefore, the cost per loaded
mile for tankers is:
88.8
clmt,flk,r (S/loaded mile) 3.08 (18)
The cost per loaded ton-mile (ctrn) for tankers is:
-------
109
3.08 88.8
cltmtanker (S/loaded ton-mile) (19)
V XV
For stoke trucks. F 569.002 and Z = 3.
The cost per loaded mile for stake trucks is:
129.38
cim take (S/loaded mile) 3.02 (20)
I . x
The cost per loaded ton-mile for stoke trucks is:
3.02 129.38
cltm (S/oaded ton-mile) (21)
see
7.3 Comparison with Actual Charges
To determine the accuracy of our costing procedure, we
compared the cost estimates using the revised costing procedure with
actual rates charged by haulers. We obtained the information on
actual rates from a study of hazardous waste haulers transportation
costs conducted by Temple, Barker and Sloane, Inc. (TBS) in May
1983 (31J.
In their cost study of drum and bulk waste transport activities,
lBS contacted a number of companies involved in the treatment,
disposal and transportation of lazardous wastes. TBS experienced
considerable difficulty in obtainin.3 cost information that could be used
to compare one operation directly to another. In fact, companies
varied in term: of type of truck, vehicle capacity, area of service,
-------
11.0
average hauling distance, quoted rates and the units to establish
rates. Nevertheless. TBS attempted to establish a uniform scale by
converting all rates to S/loaded mile.
For 5.000-6.000 gallon tankers, the quoted rates ranged from
S2.75 $4.50 per loaded mile, with an average of $3.40. Using the
average cost approach, we estimated the average cost per loaded mile
to be $4 14, which is toward the upper bound of what most shippers
are charging. However, the lower costs in the quoted range were
for one-way trips of 200-300 miles; :. listance is well above the
average one-way trip distance (84.2 miles) used in the average cost
procedure. Lising the derived cost formula for tankers with a
one-way trip distance of 300 miles, we estimated the average cost to
be 53.38 per loaded mile, which is consistent with the amount
operators reported that they charge for a 300 mile one-way trip.
For stake trucks capable of handling 70 to 88 drums, the TBS
study reported that the rate per loaded mile ranged from $2. 10 to
$4.00, with an average of $3.30. The average cost approach yielded
an estimate of $4.55. Again, the lower rates in the TBS study were
associated with longer trip lengths (200 to 300 miles) than we used.
Using the derived cost formula for stake trucks, the estimated cost
per loaded mile for a 300 mile one-way shipment is $3.45, which
compares rather favorably with the reported rates.
In conclusion, the derived cost formulas appear to be
representative of the hazardous waste transport industry quoted
rates, particularly for the long-haul market. The use of the average
-------
111
cost figures. however, should be treated more carefully, and should
only be employed when information is not available on shipment size
and trip distance.
7.4 Summary
We reviewed sev methods for e3timeting the cost of
transporting h iiL&dii wσste by trt ..k. The results varied from
gross estimates of the un cost of transport to more sophisticated
derivations of cost based on fixed and variable components. We
noted several deficiencies in these methods, particularly in the
assumptions rolattng to ihipment characteristics and the failure to
compare results to the actual rates charged by waste haulers.
We then developed a revised costing procedure which was
designed to overcome these deficiencies. Using this procedure, we
derived new cost formulas for estimating the cost of wastes
transported by tanker and stake truck. The cost estimates based on
these formulas compared quite favorably with actual industry quotes.
Cnnsequently, we feel that these formulas can be adopted for use in
policy analysis.
-------
112
CHAPTER 8
MODEL APPLICATION AND
CONCLUDING REMARKS
Below, we present two case studies which illustrate the
appitcation of the fraction release and cost medals. Th. case studiss
represent two different scenarios and demonstrate the flexibility of
the models. The results of the case studies are summarized in Table
8.1.
8.1 Scenario 1
From a policy standpoint, it is often meahingfu l to obtain
estimates of the fraction released for a large number of shipments.
Thus, we posed the following problem: Suppose 106 gallons of liquid
waste are shipped over a highway network by tanker truck. No
other information is available. What are the expected releases and
costs of transporting this material?
8.1.1 Release Computation
From Table 4.5 we used the mean distance for shipping liquids
of 77. 1 miles. Because no information was available on the nature of
the highway network, we used the appropriate mean (releasing)
accident rate of X = 2.8 x 1O accidents per truck mile from
Chapter 5. The expected amount released en route was obtained using
the fraction released from Table 5.9 as:
-------
113
Table 8.1 Summary of Results of Case Study
Sc.narlo ScenarIo 2
Quantity Shipped (gals.) i0 6 200 x 55
Distance Shipped (miles) 77.1 100
Quantity per Vehicle (gals.) 3171 2791
Average Number of Shipments 315.4 3.94
Truck Accident R t ( 1Q ) 0.28 .13
Expected Release Enrout. (gels.) 7.34 2.65
Expected Release Handling (gals.) 7.6 3.19
Total Release (gals.) 14.94 5.84
Total Release (%) 0.0015 0.053
Cost per Ton-mile CS) 0.32 0.37
Number of Ton-miles 1018.5 1180.0
Cost per Shipment (5) 325.92 429.20
Total Transport Cost (5) 102,795.17 1691.05
-------
114
E(reteased enroute) (4.2xIO 8 0.19x2.8x10 7 ) x i0 6 x 77.1
7.34 gallons
Similarly, the eipected amount released at tsrmlnal points is:
E (release at terminals) 7.6x10 6 x 1C 6
7.6 gallons
Total xpected release 14.94 gallons
8.1.2 Cost Analysis
From Table 4.5, the weighted mean shipment size for liquids is
3171 gallons, which is equivalent to 13.21 tons. Using .quation 19,
the cost per ton-mile is:
3.08 88.8
cltm (S/loaded ton-mile) = _____ _____________ 0.32
tanker 13.21 (13.21)(77.1)
Number of ton-miles. per shipment = 13.21 x 77.1 = 1018.5
Cost per shipment 1018.5 x 0.32 $325.92
Average number of shipments = 106/3171 315.4
Total Cost = 315.4 x 325.92 = $102,795.17
-------
115
8.2 Scenario 2
On a more disaggregated level, it is often useful to obtain
estimates of the anticipated fraction released for point-to-point
shipments. Thus, we formulated a problem which would be
characteristic of this class: Suppose 200 55-gallon drums are being
shipped a distance of 100 miles on Interstate highways. The ADT
and truck percentages on the highways are unknown. What are the
expected releases and costs involved?
8.2.1 Release Computation
From Chapter 5. we obtained the accident rate for Interstates
as } 0. 13 x io6 accidents per truck mile. The expected amount
released enroute was obtained using the fraction released from Table
5.9 as:
E (release enroute) = (2.4x10 6 0.lOxO.13x10 6 ) x tOO x 200 x 55
= 2.65 gallons
E (release at terminas) 2.9x1O x 200 x 55
3:19 gallons
Total expected release = 5.84 gallons
-------
116
8.2.2 Cost Analysis
The average load carried by stake trucks is 2791 gallons. which
is .quivalent to 11.6 tons. The quantity being shipped Is 11.000
gal Ions which is equivalent to 45.63 tons. Using equatIon 21. the
cost par loaded ton-ndl. is:
3.02 129.38
eltin (s/loaded ton-mile) 0.37
St. C 11.6 (100)(l1.6)
Number of ton-miles pr shipm.n 11.6 100 1160
Cost per shipm.nt 1160 x 0.37 a $429.20
Average number of shipments 3.94
Total Cost 3.94 z 429.20 $1,691.05
§.1 j jn Remarks
This project has addressed the potential risks and coits of
transporting hazardous wastes by truck. In the course of conducting
this study, we drew several conclusions that are useful for policy
analysis. Below, we briefly discuss our conclusions.
A trip profile analysis conducted on data from several states
indicated that, on average, wastes are shipped less than 100 miles
from their generation to their disposal sites. The average trip length
is lower for liquids than for solids. Generally speaking, the mean
-------
117
quantity shipped is independent of shipping distance.
In assessing truck transport risk, it is important to distinguish
between two kinds of incidents that result in spills. For one class of
incidents, the probability of occurrence is a function of the distance
traveled; for the other, the occurrence probability for a particular
shipment is fixed. We computed expected fraction release estimates
for both kinds of incidents.
The costs of transporting hazardous wastes by truck can be
reasonably approximated using the f- s derived in this study.
These cost formulas compare quite favorably with actual industry
quotes.
The individual and collective results of the entire analysis are
applicable at many levels of aggregation. Using this studys models
and cost formulas, it is possible to obtain broad estimates of expected
releases and transport costs, as well as estimatas of the risks and
costs invol,ed in individual shipments.
Perhaps the most important result of this study is that the risk
of transporting hazardous wastes by truck appears to be as large as
the potential risks at treatment and disposal sites. In fact, for some
W-E-T combinations, transport may be potentially more danηerous
activity. As a result, policymakers should give careful consideration
to the relative risks involved in the treatment, transport and disposal
of hazardous wastes.
-------
118
LIST OF REFERENCES
1. Westat Research. National Survey of Hazardous Waste
and Treatment Storage and Disoosal Facilities
und.r RCRA in 1981 , Draft Final Report, January
1984.
2. ICF Inc. t al. RCRA Risk/Cost Analysis Model : Phase
lit Report to EPATOff ice of Solid Waste JanuaF T 4.
3. U.S. EPA. Characterizations of Hazardous Wast.
Transportatiafl and Economic Impact Assessment of Hazardous
Waste transportation ulations, March 1979.
4. Ogl.sby, Clarkson H. and Gary R. Hicks, Highway
g j rin. Fourth Edition, 246 pp.
5. Massachusetts Bureau of Solid Waste Disposal, Hazardous Waste
Mana em.nt in Massachusetts, Statewide Environmental Impact
, August 1982.
6. Transportation Research Board. Trans ortation of Hazardous
Materials: Toward a. National Strategy , TR p.cia [ keport 197,
1983.
7. Transportation Research Board, Risk Assessment Processes f j
Hazardous Materials Transportation , NCHRP Report No. 103,
November
8. Rhoads, RE., An Assessment of the Risk of Transporting
Gasoline Tru . Pacific Northwest Laboratory Report
iE!i , Novemb.r, 1978.
9. Bercha, F.G. and Associates, Risks Associated with the
Transportation to Treatment of Hazardous Waste I 6stanc.s:
Phase I, Report to !nvironmental Council of Alberta,
Decembr 1980.
-------
119
10. Gaffen. C.A., An Assessment of the Risk of Transporting
Propane Truck and Train . Pacific Northwest Laboratory
Report 1 PNL-3308. March. 1980.
11. Gaylor. D.W. Statis:. . Methoth in Risk Assessment, Paper
presented at Water Pollution Control Federation. Anaheim, CA,
1978.
12. Jones. C P Barrow, R.W.. Stuckenbruck, L.C, Holt. E.L.
and Keller. h P.. Risk Analysis in Hazardous Material
Transportation Volume I Final F p EiTT Dept. of
T ansportation, Y -7 4T March, 1973.
13. Nationa Transportation Safety Board. Risk Conce ts in
Q3 s Goods Transportation tions, NTS -STS-71-1,
January. 1971.
14. Russell. E. R.. Smaltz, J J., Lambert, J. D., Delines, V. P.,
Jepsen, R.L.. Joshi, P.G. and Mansfield, T.R., Risk
Assessment Users Manual for Small Coiuiiri nity and Rural Areas ,
UT5i rtment of Transportation. R PA. Reoort
DOT/RSPA/DPB-50f81/30, October, 198i.
15. Federal Railroad Administration. Accident/Incident Bulletin No.
150, June 1982.
1 . ICF, Inc. et al. RCRA Risk/Cost Policy Model Project: Phase
2 Report TWA Office of Solid Waste, June 1982.
17. Vallett.. R., McGee, H.W., Sanders, J.H. and Enger, D.J..
Tb. Effect of Truck Size and Weight on Accident Experience
and Traffic Operations, Volume III: Accident Experience of
Trucks, FIIWA/RD-80-137, FHWA, Washington, D.C.,
July i r
18. Arthur D. Little, Inc., The Safety of High Gross Weight
Trucks , March. 1974.
-------
120
19. Federal Highway Administration, Review of Safety and Economic
Aspects of Increased Vehicle Sizes end Weights, Washingro -,
0 C.. September, 1969.
20. Bureau of Motor Carrior Safety. Federal Highway
Administration. Safety Comparison of Doubles vs
Tractor-Semitrailer Operation , Washington, D.C., November.
S
21. Zeissler. R. A Study of California Truck Accidents , CalifornL
Highway PatroC April, 1b73.
22. Scott, R.E. and ODay J., Statistical Analysis of Truck
Accident I nvol.ments DOT -HS 80C fl, N H1SA , becember,
1971.
23. Yoo. C.S., Rein, ML. and McGee, H.W., Comparison f
California Accident Rates for Single and Deub Fe Tractor-r Uer
I ation Trucks , HWA-RD-78-94THwA, March, 1978.
24. Smith, Richard N. and Edwin 1,. Wilmct Truck Accident and
Fatality Rates Calculated From California Highway Accident
Statistics for 1 ind19 i November 1982.
25. Meyers. Warren S., Comparison of Truck and Passenger-C r
Accident Rates on Limited-Access Facilities, Transportation
R.se.rch Record 808, 1981.
26. U.S. Department of Transportation, National Highway Traffic
Saf.ty Administration, Large Truck Accident Causation , DOT
HS-806 300, July 1982.
27. Booz. Allen and Hamilton, Inc. at al. Hazardous Waste
G.neration and Commercial Hazardous Waste ent
Capacity : An Assessment , Report to EPA Office of Planning
and Evaluation and Office of Solid Waste, December 1980.
28. Camp, Dresser and Mckee, Technical, Marketing and Financial
Findings for the New York State Hazardous Waste Management
Program , March i δ
-------
120
29. Arthur D. Little, Inc., A Plan for Development of Hazardous
Waste Mana sm.nt Facilities in the New Englajid gion ,
mr1.
30. Arthur D. Little. Inc., Hazardous Waste Quantities and Facility
N..ds In Maryland . August 1981.
31. T.mp .. Barker and Sloane, Inc., Survey of Transportation
Costs for Hazardous Wastes. Memo to EPA Offic, of Solid
Waste. My 18. 1983.
-------
12].
APPENDIX A
LIST OF CONTAINER TYPES
-------
122
ccJITA1I R aflR!u!IT1OIS A ll! SPCCUIC&T!ON mARI*
ABI!. OR (00 ALLY USUI4 .LY CA l l BE BlURT T Iff YEV 801 ! ll3TR COl 49 0011001h3 DESCRIPTION
SPEC ND. 00 T1 COIIT2 111*! 001T11 1 113 COIISTA CM .LED SECtION
1 YES YES 10* C M YE! 179.200 lan-Prnsvrs
2 YES YES 10* CAR YES 179.200 Han riwjr
3 YES YES TAlPI CAR YES 179.200 Nan-,mau,
4 YES YES TARE CAR YES 179.200 Nan.ψreai,as
5 115 YES 10* CAR YES . 400 Nan-,rniuyi
6 YES YES lilAC CM YE! 179400 0br,, iiu?T
7 YES YES 10* CAR YE! 179.200 llan-*vesaui
8 ES YES 111111 CM YES 179.200 Na ,r,au i,
9 ItS YES 10* CAR It! 179.200 Nairv rusige
10 YES YES TAlIC CM YES 171.200 Non riuuve
11 YES YES 10101 CAR YES 170.200 11an- rsuairs
1.2 TE l I II 10* CM YEA 170.200 Notr,rswn
13 YES YE! TANS CM TEA 179.200 Nan ,mIU7T
3.4 YES YES 10* CM YES 179.200 Ran-,maut.
YES YE! TANS CAR YES 171.200 lEm,,.u ,r,
16 YES YES 1*0 (1 CM YES 79.200 ll r,rIipiai
17 YES YES TANS CM YES 1I5.2 . Ncn.prsawve
18 115 YES T0* CM YES 579400 Nan,m.uvi
19 115 TI! lilAC CAR YES 179.200 llcir,,siuv
20 YES YES Till CM YES 179.100 P,rnvr,
21 YES YES 1*0(1 CAR YES 170.100 Pm.uvs
22 YE! YES 10* CM YES 170.100 Pmiun
23 YE! YES Till CAR YES 179. 100 Pnuura
24 TI! TI! 1* 1 1CM T15 171.100 Prmurs
25 1!! YES 1*11 CAR YE! 179.300 Nulti-wiit
26 115 . YES TAll CAR YES 170400 llulItimtt
27 YES YES TAl l CM YES 170.300 N*4ti- a4t
28 TEl YES 101 ( 1 CM 1 15 170.300 Nv1t aiit
29 015 YES 1*01 CM 1 15 170.500 11i10 , ,vrnisr
30 YES YES 1*1(1 CM YES 179.100 P,.s, ,ir,
31 111 YES 10* CAR itS 170.100 Nuoupe
32 YES YE! VJ3 L/KE0 (000 NO 19740515 573.14 Nccdan brrela and kUa (t ght
33 YEA YES BN JICtG 11000 YE! 175.154 Nooden batrels and loin (t 1nt,
34 YES YES AIYELJUA 11000 lID 19240535 171.14 Voodun barruli and loin (t lRht)
35 U YES TAI l CM YES 170.300 Nu5ti- iut
3n YES YES Till CM YES 170.300 l%uItnii tt
37 YES YES TAIl CAR YE! 179400 kcn,rrnuv,
38 ItS YES TAIlS CAR YES 179.200 11oir tsn ir,
39 fl5 YES W I CAR YES 179.200 Nsn .,rsnuro
40 115 YES TAIlS CAR 115 170.200 Noir,mnuvo
41. YES YES TAI l CM NO 19771231 179.100 Pmuun
42 itS YE! TAIl CAR NO l9fltUt 170.100 P i,uru
43 YES YES tANK CM l ID 19771Th 179.100 Preauuru
44 ItS YES lANK CM TEl 179.505 2 Prmur,
45 YES YES 10* CM YE! 179.100 2 PIIUUPA
I Sn c us no mt Pus TI BuD. ccnLaIMvI can only bs CONY! lImit Containur )
Data Sen Rttrftuti
HA2jIOT.DII5 £01111
C*JllVO.DNS cOJIT2
-------
05III61I 11 19151111CM IM VECIFCC&7101 1851523!
123
I I.e cedsi last paus
ra $ j91 e itaUI9rI aai i1v bi comi 1N .P Csnturen)
Data D i i. 1ttr b it.
NA2MT.MS 05911
C6Pl1 0JMS 05l. 12
013.01 I U1LLY I .$1LT CM 91 1 1 1.1ff mc V M I S 01111T5 tE l 49
vce rn. 05911 05192 CITIn C0W1UI CONSI! CI LLES YECILON
cOiflalIn 1310117U0N
46
523
723
TM CM
I D
179405 2 Pm ns
47
In
951
1 4 1 1CM
In
179.90023 PNuov,
Prrnuts
48
49
123
15 5
1 5$
In
14118 CM
kM CM
ID
11 5
179.105
1754052 Pr.wne
30
51
52
551
YES
151
151
5 (5
15$
1 CM 0015
l CM 0011
T4118 CM
YES
ID
5 55
179.400 UsuifIud bvdPS5r
179.400 U.vlfL.d MraNn
179.400 Lsuuif sl twdSsm
huCcelso
53
34
lE O
553
15$
ID
T CM 0115
91818 CM
151
15 8
179.400
179.400 USVIVLuC MnIsa
179.400 Usqif led I sun
55
155
955
1511 CM
15$
P,euins
56
In
In
kM C II
1
57
SB
39
60
ID
YES
TE l
YES
.-
ff1
5$
595
ID
TIM CM
TIE CM
1*91 CM
TIE CM
ItO
80
YES
YES
19771131 171400 Pmiurs
99771231179.100 h.siuvl
979.9052 Enuure
179.1052 Prawns
61
62
ES
In
155
158
IPI CM
7*91 CM
YES
YES
179.1052
179.10023 Pmcsta
175.1052 Pmwi
63
In
551
1*91CM
1 55
:9.200 IIsu-rni$w5
64
ItS
123
tIM CM
Ifs
NurrnssutS
63
10
55!
TIll CM
1 1 5
b.rr,1i m4 la s s luhsuk)
66
ID
155
)9 JUI 600 )
lID
19740213 171.94
barrels .04 km (thoU
67
YES
YES
MESELIUS r201
lID
197U*t5 179.14
I
68
15$
ID
901 flIn
111
171.210 hues
970.205 Dazes
69
YES
155
InrUES
YE!
km
70
ID
1(5
NI P1023
YES
Dazes
71
TO
YES
108 FUD
115
175.207
72
155
In
001 FITO
155
170.200
Dazes
73
115
155
ID ! P9101
ItS
178.20 ,
lazes Ilti
74
YES
lEO
I C! ElMS
115
Parer I sued ezesedid eslestarare 1
75
TO
DOS PilES
YES
970.21?
Metal lass
76
77
78
I II
ItS
YES
UI I1 TII.
11878.
101 60 Gb
555
sf8
YES
170.140
170.141 II.til Irazi
178.115 Ni 1ed
170.168 Killed
79
Its
NI 1100)
TE l
HiLled
80
115
$01 1000
515
171.969
WeLled
81
lEO
ID! 1100)
TE l
173.171 hailed
82
83
ItS
1 58
101 I X )
101 C X )
558
ItS
178.112 FIbsrbo.rd ltssd
Darn
84
YES
901 0000
YES
170.917 MetaL Used
as
10
0010000
TEl
178.190 Glued phuted v usodes boa
86
87
88
89
155
115
1(5
1 55
1006005
808 0005
551 6000
. 10160Gb
ItS
YES
ItS
115
178.911 Coodes bores I si leo f,vaeaflaz cess
173.185 P avsod ar v o4t5 bares. vir,bc d
970.984 Vsudes born. wftsbaW S
118.117 Voedus .irsbowd sveivraz
90
115
80111005
-------
fl41NER *SIIEVThTIOIG 6118 9! FICAT 1181 NWIDESS
124
1158. U!U*aT USThLLY C I I I St WU TYPE ICy DI I ! C8IIStP P 49
9!C NC. COSIt CDNT2 ClUES SIIINZR CONSIS CAIEC.LUI SECTION
tOI1TIDID. 8U IPflO11
108
109
110
1.1).
1 . 1 .2
113
1.14
11.5
1.16
1.17
118
11.9
120
121
1.22
YES 179.115
YES 178.116
ItS 173.23
YES 17 5. 117
170.118
YES 173419
1 5 5 170.193
YES 179.190
YES 175.19*
YES 179.1
NO 19790511 170.2
NO 19790511 179.3
YES 178.4
NO 19,90 U aa.,
155 179.6
Its
YES 179.14
YES 178.17
155 170.5
179.120
179. 194
t8.224
179.225
Still SIC) SMI
Steal SIC) SMM
Rucunditioned *7! (cloud *16), caws,tnd .:
1711 (own head) SIts 5
St..l SIC) ESNOS
St..I SIC) SMIS
Stul barve1.a t diuns SIC) 31 (M
Vooder, kits
Voodoo box.g, 1viaodi cleated
Vood,ii beaui hxaod , nailed
bail
lanid lead
k. kill
Saud 1.ss
lass. in r1 ood di aj
Glass, bo ilavood dr .s SIC)
Polvatbolur .. in siSal ruIn
Bins, cvsbiai.d vith e,ondabis nelislsrer.
In voodoo vivehsmd box
Gins vith s sndad .olvstvre* causer).
Saudi 5 to 6 1/2 eelian to srt lola
SIC)
Phinolirfole zuiidatad. setal saurpact.
Wooden .,otuctlvs s st
FIber dna
flbar dry ivii,ict tar ii*ldi s )aittc
contiln i ,
171.121 Fire and shock r,sistenti hinoltetus
xnlulstadi setal Ivursact.
Wooden nrot,ctiv, ovureict
Uavden Grufl, .lyimod
Uoodsn drvos . ptviiocd
Phased dn a to? slastie muds eaimtuntr
Fibetboard boxus
S ecmsl caiir .dplca l fiW buerd box far high
Fiberboard boxe
SIsal cximndsre sassluss, sex uu *20
pe avis iaWp cuaeitv
St. ,.1 cvluidir, hums, aexisy ii:. Ut
Po imd1 aatur ceracita
Iiita ljacksted
fl,ta ljscSa t ed
Nailing t tt
(eta) cans. pails end kits
171.393
179.196
170.197
179.190
*71.2)4
179410
178.219
173.30th
80 17 3 .30Th
91 YES YES
92 YES 1MIEtM.
93 935 YES LtW (ICT*0.
94 ES 8S%I* I CTIL
93 YES YES 111181 MEIOL
96 YE! D IISI IET*I.
97 913 101 11008
98 - 155 90 58008
99 YES
100 YES
101 YES CaS3OT
1.02 ItS CWOY
103 1 (5 CMDOT
104 YES C IOT
105 YE! CIR D OT
106 YES - CMIOY
107 YES CM IOY
YES CSIRIOT
YES CIESOY
YES SM CD1(TIUER YES
YES SM CONTIIIES YES
YES YES 111Th NOb-NEW. YES
YES MIII N0N-IItT& YES
YE! SAIl CONTITICS YES
YES MI t CW(TA1ICy YES
YES 88*11 NOII-fltTfd. YES
YES MUll NOIl-NETII. YES
YES 118011 I(OIMEYPL YES
It! YES DOSFIBES YES
ItS YES bOX FI YES
YES YES D01flD YES
YES YES Ttd& NO
123 YES . YES INC
124 YES CIRIOT 110 *9790511 170.8
125 YES CMBOT 110 197905*1 170.9
126 ItS 151 103! YES 179.26
1.27 YES INSIDE CWITIaIN 1 ( 5 178.20
S Si, codes on last sees
U D ,1 ) . contasne s can only ha CObOl (hair Contain,,,)
Data Ins Attubuts
11579*1.81*8 05811
WITVO.DI 1S C01T
-------
cCNT&INEA &NS&EV IATIOflS IJIS SCCIFICATIOII NWII .ERS
125
ma. OR VSJW.LY U J* ,T CAN IS bUfl TYPE lIEU D*T! CONSIR CU 49
EC NO. twill C IT2 £11 1E8 CCNT&1I CONSTI CANCwJ.DJ SECTION
CONTAThES SEStRIPTION
INSIDE CONTAIN YES
INSIDE CONTAIN YES
tNSILI CON1AIN YES
1I IIDj CONTAIN TES
INS1 , C NTA1N YES
ipim ccrnlw TE5
1115155 CONTlIN YES
HI51Dt twiTf .ii . ES
INSIDE CONTAIN YES
INSIDE CONTAIN YES
1N5155 ONTAIN YES
INSTIl l CONTAIN YES
1411( CONTAIN YES
NSTEE CONTAIN YES
U1311u5 CONTAIN ItS
INSIDE C llTAHI YES
INSIM CONTAIN YES
INSIDE CONTAIN YES
CY lINDER
YES 50 1115161.
1 55 S OZME7M .
ltS IOXI IETAI.
111 DOXICTN.
YES TAIl S
- YES - GT 1 R TES
YEA XAIAINON-IIETN. itS
Its
YES
YES
YES
YES
YES
ItS
1 53
YES
1fl. Corrsg,tid fiberboard cartons
178.23 Dw ux suer bus
171.4c Folasthyisna bottle
178.25 Metal containers aM liners
178.26 Fib,, cans aM bones
176.28 Uetir,oof suer bin s , tininsi
175.2? Pap,, Dais for lutilift
......O Linui for boise
178.31 Uotet,rsof me Unine
17 1.32 Metal cans
178.33 Nonp,f liable autal ccntoiMes
178.33a Neii ,eftllable soul containers
178.34 Ilotal tueos for padussctav, utensil
176.35 PoluI .hvlu , containers 5)0161
178.35. PulneUnilens containers 110141
178.21 Polvetlisisne CIAt3 IMPI
78. 7 FcI . .Uvlone containers
176.24 PolvetbalenD contaIners oven i gallon
casscstv 51016$
90 1fl.301h Steel cylinder. sassiess
NO 19790511 175.15 Juu in t.ΰs
YES 175.146 IIet.a1 cases. riveted or loxbieessd
YES 176.147 Metal cases. voided or riveted
YES - 178.148 115111 te.iA
YES 171.149 Metal boxes
90 173.30Th Steel cylinder, sessins. iaiiva size 10
p w.d 5 vital cuscitv
171.150 Polystyrene cases
178.19 Reusable solded ,olvsthvlSrw container
vjUtosit onei,ue 510461
1lusili s conbevi
Nont ,vsabie ..lded peivethvlsns drus for us
without viiPiCD . SIlkS
178.230 Uned cloth (triplex)
179.233 Burlap, lined
171.224 Burl. poser lined
178.131 Davis Sits 1)81
178.112 Druas STCZ 5)04* 5
170. 135 Dna; NRCS 5)81
176.137 Diana Mitt
178.130 Davis SItS 1)4*1
171.134 Steel ψverack for ansaΨ. slaatic contain s?
Mitt
1 15 175.133 Steel dims vitI polyethylene liner
NO 173.301h Steel cylinder, unless. chum si ze S
powmdi v11r esseciti
YES C4U O I 140 19790511178.12
YES 15104 NOIHCTAL YES 17L16
YES UACLOTh YES
YES 5080076 YES
155 )AOCLON YES
YES DR IIIIIETNL YES
051 541111* 1 YES
YES ONUIIIIEI&L YES
051511151*1. itS
1E5 18 1* IEDL YES
010411516%. YES
YES 05151)151*1.
YES YES 1*104
I Reproduced from
besi available copy.
128 111
1.29 YES
1.30 YES
131 ItS
1.32 155
1.33 YES
134 YES
135 ItS
136 itS
137 YES
1.38 YES
1.39 YES
140 YES
141 YES
142 TEA
143 i tS
144 ItS
143 YES
YES
YES
- YES
- YES
ES
- YES
YfY
146
1.47
148
149
1.30
131.
132
1.53
134
135
156
137
158
159
160
1.61
162
163
164
1.65
166
167
I See codes or. last sue
at Dull containers can only be CONTI (Inner Containers)
Data Base Attribute
KAZIIAT.Dit5 COIITI
c*irrio.Dcs C0N12
-------
CDrITAINE. AItBEVIAYIONE AN) !ECIS1CTION NUMBERS
126
YES 170.43 Sinless steel
YES 170.37 Sneless steel. lade of definitely erescriti
steels
YES 178.37 Seaeleu steel, cede of definitely Prescribe
steels over 1*00 rounds ate, volijn
YES 170.46 Seasltss cylinder side of definitely
prescribed alue.snus divas
Sinless steel. over 1000 n4s water volu:
Seae.less steel
hatless nickel
Seatlees steel
Seatless steel
Seethes steel
inside containers. sinless steel far WCY
YES 170.36
YES 17 2.3 5
YES 178.3 9
YES 178.40
YES 170.41
YES 170.42
YES 70.44
di
YES Sinless steel
YES Forte welded steel
NO Nonrefillible eetal containers
NO Nonrefillable eetel conteiners
ND 19301001 173.26k 1 Aleainea dries
YES 170.107 Divas
YES 170.100 Barrels or divea
YES 170.109 Umse
YES 170.136 Ora l Sits
YES 170.110 Barrels or divas Lilt
YES 170.11! Divas
YES 170.112 Drcas 5101*1
YES CYLINDER
YES CYl iNDER
YES CYLINDER
YES CYL IIW
YES E t11ME7AL
YES W IDIICYOL
YES C th1IEYAL
YES DRUI IMETAL
YES ORWMEYAL
YES DWAMETAL
YES MEYAL
YES V4UMCT*L
YES 1 1 0W! NONMETAL NO 19790511 170.10 Rubber drues
YES YES AS PAPER YES 170.236 Parer bass
YES YES BOO PAPER YES 170.237 Parer bass
YES YES BAG PAPER YES 172.2 10 Piper bass
YES YES BOO PAPER YES 170.239 Paper bass
YES YES BOO PLASTIC YES 170.241 *1 1 plastic bad
YES YES BOO CLOTH YES 170.240 Baler cloth and e r ie . lined
YES CIL INDER YES 178.49 Forte welded steel
YES CY l iNDER YES 170.56 Welded steel
YES - CYLINDER YES 170.50 Welded and bra:ed steel
YES C YUNC I ER YES 178.15 Welded and brazed
YES CYLINDER YES 170.54 Welded or welded and brazed
YES . CYLINDER NO l73.304d 3 Cylinder without Iontitudir.a l ccci for
pressures of 150 to 500 rounds rsi
206 YES CYLINDER YES 178.51 Welded or brazed steel, cede of definitely
prescribed steels
See codes on last eec sa Bulb. containers can c ml v be CC!4T1 (Imer Containers)
Data Bese Attrtbuti
12?1. a; U!aLY USaLY CA !! BC BIU5I TYPE NEW OAT! CONSYR CR 49 CONTAINER DESCRIFTION
S!EC k3. CINTI CONY? EITHER CONTAINER CONSYR CAC LED SECTION
178.65 Nanrivaible (non-reftllable) calirgere ND:
170.36 hatless steel
168 YES CYLINDER YES
169 YES YES CYLINLERB1LX YES
170 YES CY l iNDER
171 TES CYLINDER
172 Y ! - YES CYLINIeEI ITS.
173 YES CYL1NDER
174 YES YES CYLINDER Yt
175 YES CYLI I IDER
176 YES CYL IIrvER
177 YES CYLI I IDE R
178 YES CYLINDER
179 YES CYLINDER
180 YES CYL INDER
181
182
183
184
185
186
187
188
189
190
191
192
193
1.94
195
196
197
198
199
200
201
202
203
204
205
178.45
178.48
NAZMAT.DSS tONIl
CMTWO.DMS tDJIT2
-------
t lT*1WR e)$lZVIoTIWiS ID EC1rI1AT1OI1 MJ ! B4 8
127
6614. 86 U!USUj 83VN.LT C86 I L 813.51 1 1171 WV DATE COIiSTR R 49
57CC k. CO T1 IlT2 CIIWI CWIYAIII3$ CO41# CMrJLLD SECTION
C TAIN!R IEStEIflION
201 IU
208 ItS
209 itS
210 133
211 itS
212 itS
213 T
224 113 YES
215 YES
216 itS
217 itS
218 YES
219 itS
220 YES
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
YES
113 -
It -
1 1 5 115
YES -
YES It!
YES -
its - -
YES - -
YES -
Its
its - -
1 5 - -
YES - -
YES 113
YES itS
YES TEE
YES
CtII0 YES 173.61 Vold.d otto!
CIttMDES YES 176.52 $.ld.6 end buu4 steel
CTLUI $8 3 1 3 5 173.53 Inosda rentiinen , welded it,,!
yts i e.si lneiψo eontzinirs. u, le, st.. to, u sa
CtT. IN OCR YES 173.47 Inuode contauierie weldid etsuiI,u, sloe
CTUHDCR YE! 173.68 1414.4 aIwetnus
CIUNILS YES 17L17 ll.ld.d . uisulst*d
tNJM Wild. YES 17L60 Steel barreli or dr ..s f6 1*1
11$ 1 10 1fl.32c St.,! sortable tvl
its YNIE itS 171.245 Slit!
ItS 16W 10 173.3 Slit! portable tail
TE$ 11W NO 112.324 Ilvainus or iaenesius portable t *i
115 TN NO 173.324 Celindrlcal alumina sortable tail.
RAN CCNTAIWR 110 0750351112.313*2 Getil erciud, uranius or lead .k(.lded
tontainer for radssaetiv aitiriali
113 11W YES 178.352 lIt.!
113 INN YES .253 Metal
ENOJI WIlL YE! 178.81 Stool borrels or drumi 061*3
86tj1 WIlL YES 173.10 Stool barr60i or 4r as 01*6
86131 WIlL YES 178.83 Stool birrels or dvuss 0 1 1*3
61 INTM. ItS 173.64 St. ,) Ditrels or 4r as. lined £301
8610 WIlL itS 178.85 Sti. ! drums 0101
M WIN. YES 178.87 Stitl borrels or dras R..d lined RI0 Z
00* INTAl. YES 178.88 NIckel barrels or drums 1)046*
1*131 WIN. YES 178.69 Stool barrels or drums 0*1 1
0051 WIN. YES 176.90 Mane! drums
1mM WIN. ItS 173.92 LI611d stool drums £0112
0051 hEll!. YES 178.91 5t. l drums abasran lined 01412
13$ 1N YES 176.253 5th)
1I Will. 115 173.97 Steel bat .1. 0 , drumS Oil.?
D Rill WIN. YES 176.9$ St..! barrels or drums 04*3
Dl i i i WIN. YES 178.93 St ..l barrels or drums 046*
10411 WIN. fE! 178.102 CSliftdr%Cil Otto! OtoIPi lJ strught uidi i
for inside elastic cantaintert
0051 ICIAL 135 173.100 Stool barrels or drums 134*1
01)31 WIN. YES 173.101 Stool barrels or drums PillS
812 CON1AIWI YES 173.103 lieU! eactalilo
RAN CO3ITAIl YES 179.104 IleUl ,at&I I IM
NAN CCNTAIICR 115 178.350 General packalirO , for tv, A r;4ioactav,
s.t,ria ls
C YI.1I013 YES 178.59 Stool for acttvleae
C fl.1ND YES 173.60 Steel for ,cetthn.
C1U35CR it s Ilonr,ftllabli metal cantainers
3*6 .DTh YES Cloth or burla b o O fcvntl for aolid
isterisli)
5$ lull containers can only be COIflI lltii ,v Containers)
Data Base Attribute
lffihIIA1.DlI5 .C21fl1
C*lfTV .MS Wfl
239 YES 15
240 its YES
241 Its
242 itS
243 TES
244 ItS
245 YES
246 ItS
247 YES YES
codeS on loot cal l
-------
1.28
C0I1AII R &1112U 1&YIONS MB QICIflCS.tlCK PItIlESS
ten. e Ltsu.Y IYSVYLLY W I % IZJ IE lift 10 M IS CCiflE i cannio pss ini .
SPEC C. CCIII C21&12 titlISt CCIflAINES CCXSIR foCfl4fl SECT ION
248 YES YES SAC PUflIC YES Plutte ba tewit l for islid ostortalol
249 ItS YES SAG PAPER YES Pry In Ietrtl for oolid ostino lol
250 YES OTIG YES Pactauo or cortuonun crioflod a bosS a
o l i n;? I Iv u rotOords? on udlitta i i
1 4 k M
251. YES YES ODD YES tons lug, sMu If till) ocairoud Sum
loMliw or %mjoodini l
252 ItS YES PI EIJbII vO i YES ilissir burnt Itust l fit oiled outurtolul
253 ItS YES CTIJWVES flfl C 1?270701 tlS3lo I Cyliuduru t I C t.i 2500 owudu atom nlaO
kilL 1RMIP 1 AT
254 TEl YES ODES YES Portal, Ion (esiti for joild ostertulol
255 YES 01152 YES Snorter lift catsorwi blurt
256 YES l OT tE YES bottlur butte or o li n itt oucofoud.
eurseoti 2 ulla or ins
257 YES 1071 1 1 YES Glass bottlou coxocity 2 ulIm ur Ins
2,8 YES SOTSU YES Platte bottlur toronto 2 oub ino or trio
259 YES YES SO S I L Sour vat or ftbultoord not inciftud
260 TEE YES POT P153 YES Fiberboard bus or aria
26 1. YES YES SOXPETS. 13
262 YES YES 105 5005 YES Pooduebox
263 YES 07 10 13 tnt nude of icodurs tutu uoUi vI M eat
Ceonti urn ,)
264 YES YES C S YES turn oilier th a i suit) or ilmino
265 YES CS YES Aerosol ciel teontinto 1Mg ? nussurtl
266 YES CAN YES Alunruse in
267 YES YES t N t YES F ob ur boo r dca
268 YES YES CAN YES Sit u corn cruelty 7 uullurro or loot
269 YES tASt IlY YES CurWg,otherthin lbi u lo? ?list lCor
osturis l wnonnftodr cnantu 5 islboun or
ton
270 YES t A SSOY YES Gino turbo s, cosocttv Z ealLons or tori
271 YES CMflY YES P las m c iiiss cuncoty 5 suilirno or tot ,
272 YES YES CONTIHO YES Plaotar aria , or box Iccr42 orison);)
27) YES YES CoIuThbPB YES Containun mc doocrtrtion sivor (do itt vi i
of it oil posumblo)
274 YES INSIDE cmn&lii YES Shut conta iner , it carcotu or doicrutlor
liven ,,
275 YES SM CONTASIER YES Lad contour used is sluo ldtr.u for wiur
container of ,odtogttovi aterouls
276 YES INSIDE CONIES YES Ploottv contaorcr. it ctrscttv or dmrart:
its ,
277 YES O IlIER YES Roltud stytofoso ouurioci br tuttirsi ajar
or corbous
278 YES CYI.IWVER YES Cylinder a orussoru store ! for co r i n t H
Iii
I Si, cc i i i urn lost iii , to Sod t contstnurs con curly be CONY) linuor Containers)
Datu Sue Attnbutu
IIAVIAY.IIIIS 00511
cara.uics n
-------
129
NIIi14& 815151116710115 aIflP S15CUIEATIO1I 1101 1)ERS
;UP.. L!U& .LY EdJ.L1 C l ii !E WJ.i$ TYPE l 1I 5815 COIISTR R 41 CONT4INES IESCEIflIOU
tC Lt. COIII I CCi 1T2 (111189(0111811155 (011315 CAICflL (D 58(11011
279 185 0TI R YES Colindricil astil cinteiner. net for
couv.ilu Sails I i. . net a pmsjrv
vs i as l)
280 YES YES 5102 1 185 hr m - fiber, s,taL or .hstic not ovecif is o
281 153 1 ( 5 56121 IOM tlL 755 ftbn dnm conti for selson. cont2 for
Usuidi
282 155 ES 56 JII TfJ . 153 ItetaidNIa
283 YES 56151 IOIICfli. 185 Plutoc dr .
284 ItS 5612% Il0II- 1581N . YES 10* 1st Orra
285 ItS 0fl 185 5IAe op ;rvn ti . t for tIe iliitslnt of
,curv
286 YES YES N PU 1 83 Refi beset car for solid u sna1s u niv
287 185 185 I FU 155 HilIwia boner trader for solid utsvisls
288 155 YES CIU1IIIO 511.11 P61 11270701 173.31. 2 Ciiindsr. 170* tvatds vets? vol .ac FCc IAIL
T6NISPOIT OILY
289 185 YES 1N flIYESI10NL 155 17t.271 St.ei .ortabls link
290 155 785 INS. INTESICN . 155 175.V2 Steel verteble U
291 155 . 1 8 8 It S V hess. rustic or earthorvars not
i ,cified
292 155 JIS. 185 G1an u
293 155 155 P l a stu. .er
294 YES P i G YES Jul. liii. Ot pirstie not peetfi,d, catscitt
lore IJIA 2 gallOns lid lii i than 5 1.11cn$
295 YES YES Gins cavacitv sore tMr. 3 nuns end
less than S gallons
296 155 1 (5 Plastic Pug, opacity sore than 2 siflon, n
len than 5 hillOll.
297 1!! Kt 5I TM. YES IIvti l S .fl
298 755 785 IIESEJEEO 1 10(0 185 Joo4.n ku
299 YES INSI St (561*111 7 ( 5 Plastic liner for fiber Irons and boxes or
setal Irons eontau iri lioulds
300 155 15 5 0110 1 8 5 auInhir Winlo in bus or airerift
301 YES 0110 155 175.315 For hevid nitrohlvcurtn of distblsns sine.
dinitreti
302 YES 0710 YES 175.315 Ccntnnc? for blestini can
303 188 155 1885 P10 173.33 (;r %arlj
304 YES 15! TNS NO 173.33 Cm i tanks
305 ItS 155 T81 NO 173.33 Carlo tanks
306 185 YES 1N8 110 173.33 Carlo tarki
307 1 (5 YES TfJ NO 173.33 Cuss teruks
308 155 7611$ NO 173.85 Carlo tanks
309 158 155 1611$ 155 179.341 Curio tanks
310 158 YES TIllS 1(3 170.342 Carlo terM
311 153 1 (5 1611$ 110 173.33 Carl, tails
* See codes or. Last sea, U Bulk cψfttithers can only Is (01 (11 fInest Containers)
Date less 6ttributo
11*2581.585 CDIITI
-------
CGNTI4IIER ABMEI8AYIONS AtE VEC ITICAYIOK WJISBESS
130
em.. ottkSThU.TUS%3*3.YCk$$t tiLts TYPE NEW B4ltCtNStt as is
SPEC It. CGNT1 CONT2 cilia CONTAINa - COMETS CMCZLLD SE C T ION
COMYSI)O IESCflPlIOn
312 YES - YES liME
33.3 TEE TEE T IME
314 YES YES YW1
315 YES ItS YAMS
33.6 YES YES Till. CRY?
317 YES
315 YES TE l PAIL
319 YES YES I IIINE1AL
322 YES YES TANS
323 TEE YES T i l l CM
324 YES YES T ilt
325 YES YES Till
326 YES TEE YAMS
327 itS TEE TAJI I
325 YES YES t i llS
NO 173.33 !ifls tins
YES 175.343 Cuss tarn
em im. Cites tails
YES 175.337 Cans tails
YES Cares tanks fir cnossriiC hinds
YES kited m batten resists uh i l. rerestit stated
1W tIftUIM used
TEE Pat h uses hesdi ened ls 10 ashiest or lees
YES Metal set h pen heads tai nts 10 salient us
ins
Plastic call. vai l heads tssacits 10 SaIlors
or less
flints uses esila for bitters rporti sties :
other caistalser swr
YES Nmsortable tail
YES Sssssosd tai l. ts r
YES Portable tail
TEE Portable She, tail
YES Statue tfl
YES Tail tricks task. sstmted en trick chassis
YES Tail. tinhers enstrailer it full trailer
CUss axles)
Saunas tube
FlWr tsar
Discs toW
Its4sna toWs Isberbosrd
173.3 53b test A container fur redisict lve saterishi
173.395c Test P containers for rsd sstttvs
saterial itncludes sisAl PUtties thru Attic
casks)
320 YES DEUISIICN-ICEYAL YES
321 YE! e l s a YE!
329 Y15 TUBE YES
330 YES YES TU BE us
331 YES TUBE YES
332 YES YES TUBE YES
333 YES l iii CONTAINER YES
334 YES YES Priit CONTAINER YES
I Ste codes on lest sate
it Sulk containers ten only be CONY! (Inner Containers)
Data Bass Attribute
NAVIAT.DtIS ttNll
CAJiT VC. vr.s CONY2
-------
131
1! USED III LL . JCIiItM 611011W!
cODE DCE cS1P11 l
I Iii,Ui rp I1 .(MmS .4I)
2 Ie1 Hse?rnu Container
3 0uiI Red Container
4 Du1l Irteneodel Container
5 hiD Vatir Container
6
7
cODES USED Ill RESTRICTILCCX 6YTRIDUT !
COO! &WEVTAT2 I DCSt&IPTI
1 mia Rueovable lead bitliorizid
2 Rt tenovibis lead Resulted
3 D 4 !uevabls lead Nat AuU odzad
4 l C Nen-Reuenble Container
S SIC SinIli Tn. Container
6 Ut Tot Aircraft Oil
J 5tC- 46 Smile Tn. Container I
Ruovabie hid *uthorl&ed
S STC-RIU 6 Sinolu Tn, Container I
luiev* Ie Head Not Mlmrlzad
9 STC- Sinus In. Container $
Re.ovab le lied IN IDId
10 NSC4IA Kon- r iabls Container I
Recovable Heed Nothorised
II NRC-RIO* Nofi4eueabli Container I
Risovable Haid Hot gUmr1zid
12 hTt-RI NanAsusable Container $
Re,cvable heed Rievirid
r oo ii: cairsi.x.orc
CODES USED IN TTPLOJECORD 611011W!
DESCRIPTION
I Genetic container nine isid yuan no
s ecsflcition ccntainir is liven
2 014 sneciflcatian c tunin. contlsoad
use alloyed, no new conitiution
3 . ..t 0.0.1. ipeclfication container
faayu i l 49 Fart 170
4 Old snecificatlon container, no long.,
guthsni:.d Per ha:undsus saterlalu
5 Provesid siucificition container
A Perfonsanica v.ciflcation for
redicactive saterul container
7 Sesclflcatien converted during
reconditionini. 17 1 /27W
FR produced from
[ best ev&Iablo copy .
-------
132
APPENDIX B
DESCRIPTION OF FAILURE MODES AND CAUSE CODES
-------
133
FAILURE MODES
Cod.
Numb.r
Abbr.vlatlon
01 DROPPED
02 EXT PUWCT
03 OTHER FRt
04 WATER
05 OTHER LIQ
06 FREEZING
07 EXT HEAT
08 INT PRESS
09 CORR-RUSt
10 DEF FVC
11 LOOSE FVC
12 INNER REC
13 BOTTOM
14 BODY-SIDE
15 WELD
16 CHiME
17 OTHER
18 HOSE BUST
19 LOAD-UNLD
20 IMP BLOCK
21 IMP LOAD
22 VEH ACC
23 VENTING
24 FUMES
25 FRICTION
26 STAT ELEC
27 METAL FTG
D.scrnpt lon
Dropped in Handling
External puncture
bamage by other freight
Water Dama
Damage from other liquid
Freezing
External heat
Internal pressure
Corrosion or rust
Defective fittings, valves or
dosu res
Loose fittIna valves or
closures
Failure of inner receptacl.s
Bottom failure
Body or side failure
Weld failure
Chime failure
Other conditions
Hose burst during
loading/unloading of tank trucks
Loading/unloading spill
(involving tank trucks and
trailers)
Improper blocking/bracing (cargo
shifted, fell over, etc.)
Improper loading (upside down,
on the side, heavy freight on
top)
Vehicular accident or derailment
Venting (automatic or intentional
manual venting)
Release of fumes only (any type
of container)
Friction (between containers or
containers and vehicle)
Static electricity
Metal fatigue
-------
CAUSE CODES
Cod. Numb.r D.script on
01 Human error
02 Package failure
03 Vehicular iccidents
04 Other
134
-------
135
APPENDIX C
INCIDENT FREQUENCY AND DAMAGE HISTOGRAMS
-------
FREOUENCT CONTAINER CLASS I
136
50
30
U
z
20
10
0
15 17 19 22 211
1 2 3 7 8 9 10 11
FAILURE COCE
-------
137
C0NTR1NE CLASS 3
9000
6000
7000
6000
E500 0
w
&0o0
3000
2000
1 000
0
1 2 3 7 6 9 10 11 33 14 15 17 19 22 211
FAILURE CODE
-------
138
FREQUENCY - CONTAINER CLASS 2
300
200
U
z
U
U
100
9 . . - . nnH .
1 2 3 5 6 7 6 9 1O11 2I31II1S17l92O212227
FFIILURE CODE
-------
139
- CONTAINEJi CLR3S 2
70000
60000
50000
w
w
3000o
20000
10000
0 ] . I I -
I 235691O11l2L3Il l5l6l7l9202l2227
FRILURE CODE
-------
140
FREQUENCY - CONTRZNER CLRSS 3
200
150
z
100
I ii
I
50
o - - - .
1 2 3 1 6 8 10 11 12 13 II I 17 19 20 21 22 27
FRILURE COOE
-------
141
- CONTAINER CLASS 3
20000
15000
u 100O0
I ,
5000
0
123146
13 Ill 17 19 20 21 22 27
CODE
-------
142
FREQUENCY CONTAINER CLASS 14
1400
300 [ 1
U
z
w
0
100
0 - - n. . - - - H R .
1 2 5 14 5 6 7 8 9 10111213114 161719a02122
FAILURE CODE
-------
143
RMAGC - CONTAINER CLASS 1
soO
100
0
z
100
0 flfl . -.
1 2 3 II 5 6 7 8 9 I0)112131Ill6 7192O2122
FAILURE CODE
-------
144
F COU(NCT CONTAINCA CLASS 5
2S0
200
ISO
U
2
U
a
I I
100
a
50
o . . . flflHfl . . .
1 2 3 II 5 7 8 9 10111213111516 1719202122
FAILURE CODE
-------
145
άItRC( - CONTAINER CLASS $
300 - -
250
200
C
a
a
0
£
. 150
50
o ] _____ __ I
1 2 3 II 5 7 8 9 10111213111 1516 1719202122
FAILURE CODE
-------
146
FAEOUENCY - CONTRINER CLR5S 8
500
400
300
2
11.1
a
taJ
2DO
100
FRILURE CODE
-------
147
- CONTAINER CLASS 6
5000
4000
.5
2O00
1000
0 _._.. .n_nHHn_
1 2 3 6 7 8 9 10 It III 15 17 18 19 21 22 2 1
FAILURE CODE
-------
148
FR 0UENCY - CONTAINEA CLASS 7
1500
1250
1000
3
750
500
250
0 U r, HHrI. flH r rir
1 2 3 5 7 8 9 1011 121311115161719202 12227
FAILURE CODE
-------
149
- C0NT 1NEA CLRSS 7
1200
900
a
C
0
S
0
600
13
0
300
o fl 1s. fl _
1 2 3 5 7 8 9 lOt1l21 1t115 1617192021 2227
FRILURE CODE
-------
150
FREQUENCY CONTAINER CLASS a
300
200
I . )
z
(U
0
100
0
FAILURE CODE
-------
151
- CONTRIHER Ct.R5S 8
30000
25000
20000
15000
10000
5000
0
1 2 3 5 7 8 9 101%12t3t1U516t7t920212227
FR1UJF E CODE
------- |